>>> 2024-01-06 usb on the go

USB, the Universal Serial Bus, was first released in 1996. It did not achieve widespread adoption until some years later; for most of the '90s RS-232-ish serial and its awkward sibling the parallel port were the norm for external peripheral. It's sort of surprising that USB didn't take off faster, considering the significant advantages it had over conventional serial. Most significantly, USB was self-configuring: when you plugged a device into a host, a negotiation was performed to detect a configuration supported by both ends. No more decoding labels like 9600 8N1 and then trying both flow control modes!

There are some significant architectural differences between USB and conventional serial that come out of autoconfiguration. Serial ports had no real sense of which end was which. Terms like DTE and DCE were sometimes used, but they were a holdover from the far more prescriptive genuine RS-232 standard (which PCs and most peripherals did not follow) and often inconsistently applied by manufacturers. All that really mattered to a serial connection is that one device's TX pin went to the other device's RX pin, and vice versa. The real differentiation between DCE and DTE was the placement of these pins: in principle, a computer would have them one way around, and a peripheral the other way around. This meant that a straight-through cable would result in a crossed-over configuration, as expected.

In practice, plenty of peripherals used the same DE-9 wiring convention as PCs, and sometimes you wanted to connect two PCs to each other. Some peripherals used 8p8c modular jacks, some peripherals used real RS-232 connectors, and some peripherals used monstrosities that could only have emerged from the nightmares of their creators. The TX pin often ended up connected to the TX pin and vice versa. This did not work. The solution, as we so often see in networking, was a special cable that crossed over the TX and RX wires within the cable (or adapter). For historical reasons this was referred to as a null modem cable.

One of the other things that was not well standardized with serial connections was the gender of the connectors. Even when both ends features the PC-standard DE-9, there was some inconsistency over the gender of the connectors on the devices and on the cable. Most people who interact with serial with any regularity probably have a small assortment of "gender changers" and null-modem shims in their junk drawer. Sometimes you can figure out the correct configuration from device manuals (the best manuals provide a full pinout), but often you end up guessing, stringing together adapters until the genders fit and then trying with and without a null modem adapter.

You will notice that we rarely go through this exercise today. For that we can thank USB's very prescriptive standards for connectors on devices and cables. The USB standard specifies three basic connectors, A, B, and C. There are variants of some connectors, mostly for size (mini-B, micro-B, even a less commonly used mini-A and micro-A). For the moment, we will ignore C, which came along later and massively complicated the situation. Until 2014, there was only A and B. Hosts had A, and devices had B.

Yes, USB fundamentally employs a host-device architecture. When you connect two things with USB, one is the host, and the other is the device. This differentiation is important, not just for the cable, but for the protocol itself. USB prior to 3, for example, does not feature interrupts. The host must poll the device for new data. The host also has responsibility for enumeration of devices to facilitate autoconfiguration, and for flow control throughout a tree of USB devices.

This architecture makes perfect sense for USB's original 1990s use-case of connecting peripherals (like mice) to hosts (like PCs). In fact, it worked so well that once USB1.1 addressed some key limitations it became completely ubiquitous. Microsoft used the term "legacy-free PC" to identify a new generation of PCs at the very end of the '90s and early '00s. While there were multiple criteria for the designation, the most visible to users was the elimination of multiple traditional ports (like the game port! remember those!) in favor of USB.

Times change, and so do interconnects. The consumer electronics industry made leaps and bounds during the '00s and "peripheral" devices became increasingly sophisticated. The introduction of portables running sophisticated operating systems pushed the host-device model to a breaking point. It is, of course, tempting to talk about this revolution in the context of the iPhone. I never had an iPhone though, so the history of the iDevice doesn't have quite the romance to me that it has to so many in this space [1]. Instead, let's talk about Nokia. If there is a Windows XP to Apple's iPhone, it's probably Nokia. They tried so hard, and got so far, but [...].

The Nokia 770 Internet Tablet was not by any means the first tablet computer, but it was definitely a notable early example. Introduced in 2005, it premiered the Linux-based Maemo operating system beloved by Nokia fans until iOS and Android killed it off in the 2010s. The N770 was one of the first devices to fall into a new niche: with a 4" touchscreen and OMAP/ARM SoC, it wasn't exactly a "computer" in the consumer sense. It was more like a peripheral, something that you would connect to your computer in order to load it up with your favorite MP3s. But it also ran a complete general-purpose operating system. The software was perfectly capable of using peripherals itself, and MP3s were big when you were storing them on MMC. Shouldn't you be able to connect your N770 to a USB storage device and nominate even more MP3s as favorites?

Obviously Linux had mainline USB mass storage support in 2005, and by extension Maemo did. The problem was USB itself. The most common use case for USB on the N770 was as a peripheral, and so it featured a type-B device connector. It was not permitted to act as a host. In fact, every PDA/tablet/smartphone type device with sophisticated enough software to support USB peripherals would encounter the exact same problem. Fortunately, it was addressed by a supplement to the USB 2.0 specification released in 2001.

The N770 did not follow the supplement. That makes it fun to talk about, both because it is weird and because it is an illustrative example of the problems that need to be solved.

The N770 featured an unusual USB transceiver on its SoC, seemingly unique to Nokia and called "Tahvo." The Tahvo controller exposed an interface (via sysfs in the Linux driver) that allowed the system to toggle it between device mode (its normal configuration) and host mode. This worked well enough with Maemo's user interface, but host mode had a major limitation. The N770 wouldn't provide power on the USB port; it didn't have the necessary electrical components. Instead, a special adapter cable was needed to provide 5v power from an alternate source.

So there are several challenges for a USB device to operate as host or device:

• The USB controller needs a way to determine if it should behave in host or device mode. Ideally, the user shouldn't have to think about this.
• The USB controller needs to be able to supply power when in host mode, and in most practical situations also needs to accept power (e.g. for charging) when in device mode.

Note that "special cable" involved in host mode for the N770. You might think this was the ugliest part of the situation. You're not wrong, but it's also not really the hack. For many years to follow, the proper solution to this problem would also involve a special cable.

As I mentioned, since 2001 there has been a supplement USB specification called USB On-The-Go, commonly referred to as USB OTG, perhaps because On-The-Go is an insufferably early '00s name. It reminds me of, okay, here goes a full-on anecdote.

Anecdote

I attended an alternative middle school in Portland that is today known as the Sunnyside Environmental School. I could tell any number of stories about the bizarre goings-on at this school that you would scarcely believe, but it also had its merits. One of them, which I think actually came from the broader school district, was a program in which eighth graders were encouraged to "job shadow" someone in a profession they were interested in pursuing. By good fortune, a friend's father was an electrical engineer employed at Intel's Jones Farm campus, and agreed to be my host. I had actually been to Jones Farm a number of times on account of various extracurricular programs (in that era, essentially every STEM program in the Pacific Northwest operated on the largess of either Intel or Boeing, if not both). This was different, though: this guy had a row of engraved brass patent awards lining his cubicle wall and showed me through labs where technicians tinkered with prototype hardware. Foreshadowing a concerning later trend in my career, though, the part that stuck with me most was the meetings. I attended meetings, including one where this engineering team was reporting to leadership on the status of a few of their projects.

I am no doubt primed to make this comparison by the mediocre movie I watched last night, but I have to describe the experience as Wonka-esque. These EEs demonstrated a series of magical hardware prototypes to some partners from another company. Each was more impressive than the last. It felt like I was seeing the future in the making.

My host demonstrated his pet project, a bar that contained an array of microphones and used DSP methods to compare the audio from each and directionalize the source of sounds. This could be used for a sophisticated form of noise canceling in which sound coming from an off-axis direction could be subtracted, leaving only the voice of the speaker. If this sounds sort of unremarkable, that is perhaps a reflection of its success, as the same basic concept is now implemented in just about every laptop on the market. Back then, when the N770 was a new release, it was challenging to make work and my host explained that the software behind it usually crashed before he finished the demo, and sometimes it turned the output into a high pitched whine and he hadn't quite figured out why yet. I suppose that meeting was lucky.

But that's an aside. A long presentation, and then debate skeptical execs, revolved around a new generation of ultramobile devices that Intel envisioned. One, which I got to handle a prototype of, would eventually become the Intel Core Medical Tablet. It featured chunky, colorful design that is clearly of the same vintage as the OLPC. It was durable enough to stand on, which a lab technician demonstrated with delight (my host, I suspect tired of this feat, picked up some sort of lab interface and dryly remarked that he could probably stand on it too). The Core Medical Tablet shared another trait with the OLPC: the kind of failure that leaves no impact on the world but a big footprint at recyclers. Years later, as an intern at Free Geek, I would come across at least a dozen.

Another facet of this program, though, was the Mobile Metro. The Metro was a new category of subnotebook, not just small but thin. A period article compares its 18mm profile to the somewhat thinner Motorola Razr, another product with an outsize representation in the Free Geek Thrift Store. Intel staff were confident that it would appeal to a new mobile workforce, road warriors working from cars and coffee shops. The Mobile Metro featured SideShow, a small e-ink display (in fact, I believe, a full Windows Mobile system) on the outside of a case that could show notifications and media controls.

The Mobile Metro was developed around the same time as the Classmate PC, but seems to have been even less successful. It was still in the conceptual stages when I heard of it. It was announced, to great fanfare, in 2007. I don't think it ever went into production. It had WiMax. It had inductive charging. It only had one USB port. It was, in retrospect, prescient in many ways both good and bad.

The point of this anecdote, besides digging up middle school memories while attempting to keep others well suppressed, is that the mid-2000s were an unsettled time in mobile computing. The technology was starting to enable practical compact devices, but manufacturers weren't really sure how people would use them. Some innovations were hits (thin form factors). Some were absolute misses (SideShow). Some we got stuck with (not enough USB ports).

End of anecdote

As far as I can tell, USB OTG wasn't common on devices until it started to appear on Android smartphones in the early 2010s. Android gained OTG support in 3.1 (codenamed Honeycomb, 2011), and it quickly appeared in higher-end devices. Now OTG support seems nearly universal for Android devices; I'm sure there are lower-end products where it doesn't work but I haven't yet encountered one. Android OTG support is even admirably complete. If you have an Android phone, amuse yourself sometime by plugging a hub into it, and then a keyboard and mouse. Android support for desktop input peripherals is actually very good and operating mobile apps with an MX Pro mouse is an entertaining and somewhat surreal experience. On the second smartphone I owned, I hazily think a Samsung in 2012-2013, I used to take notes with a USB keyboard.

iOS doesn't seem to have sprouted user-exposed OTG support until the iPhone 12, although it seems like earlier versions probably had hardware support that wasn't exposed by the OS. I could be wrong about this; I can't find a straightforward answer in Apple documentation. The Apple Community Forums seem to be... I'll just say "below average." iPads seem to have gotten OTG support a lot earlier than the iPhone despite using the same connector, making the situation rather confusing. This comports with my general understanding of iOS, though, from working with bluetooth devices: Apple is very conservative about hardware peripheral support in iOS, and so it's typical for iOS to be well behind Android in this regard for purely software reasons. Ask me about how this has impacted the Point of Sale market. It's not positive.

But how does OTG work? Remember, USB specifies that hosts must have an A connector, and devices a B connector. Most smartphones, besides Apple products and before USB-C, sported a micro-B connector as expected. How OTG?

The OTG specification decouples, to some extent, the roles of A/B connector, power supply, and host/device role. A device with USB OTG support should feature a type AB socket that accommodates either an A or a B plug. Type AB is only defined for the mini and micro sizes, typically used on portable devices. The A or B connectors are differentiated not only by the shape of their shells (preventing a type-A plug being inserted into a B-only socket), but also electrically. The observant among you may have noticed that mini and micro B sockets and plugs feature five pins, while USB2.0 only uses four. This is the purpose of the fifth pin: differentiation of type A and B plugs.

In a mini or micro type B plug, the fifth pin is floating (disconnected). In a mini or micro type A plug, it is connected to the ground pin. When you insert a plug into a type AB socket, the controller checks for connectivity between the fifth pin (called the ID pin) and the ground. If connectivity is present, the controller knows that it must act as an OTG A-device---it is on the "A" end of the connection. If there is no continuity, the more common case, the controller will act as an OTG B-device, a plain old USB device [2].

The OTG A-device is always responsible for supplying 5v power (see exception in [2]). By default, the A-device also acts as the host. This provides a basically complete solution for the most common OTG use-case: connecting a peripheral like a flash drive to your phone. The connector you plug into your phone identifies itself as an A connector via the ID pin, and your phone thus knows that it must supply power and act as host. The flash drive doesn't need to know anything about this, it has a B connection and acts as a device as usual. This simple case only became confusing when you consider a few flash drives sold specifically for use with phones that had a micro-A connector right on them. These were weird and I don't like them.

In the more common situation, though, you would use a dongle: a special cable. A typical OTG cable, which were actually included in the package with enough Android phones of the era that I have a couple in a drawer without having ever purchased one, provides a micro-A connector on one end and a full-size A socket on the other. With this adapter, you can plug any USB device into your phone with a standard USB cable.

Here's an odd case, though. What if you plug two OTG devices into each other? USB has always had this sort of odd edge-case. Some of you may remember "USB link cables," which don't really have a technical name but tend to get called Laplink cables after a popular vendor. Best Buy and Circuit City used to be lousy with these things, mostly marketed to people who had bought a new computer and wanted to transfer their files. A special USB cable had two A connectors, which might create the appearance that it connected two hosts, but in fact the cable (usually a chunky bit in the middle) acted as two devices to connect to two different hosts. The details of how these actually worked varied from product to product, but the short version is "it was proprietary." Most of them didn't work unless you found the software that came with them, but there are some pseudo-standard controllers supported out of the box by Windows or Linux. I would strongly suggest that you protect your mental state by not trying to use one.

OTG set out to address this problem more completely. First, it's important to understand that this in no way poses an exception to the rule that a USB connection has an A end and a B end. A USB cable you use to connect two phones together might, at first glance, appear to be B-B. But, if you inspect closer, you will find that one end is mini or micro A, and the other is mini or micro B. You may have to look close, the micro connectors in particular have a similar shell!

If you are anything like me, you are most likely to have encountered such a cable in the box with a TI-84+. These calculators had a type AB connector and came with a micro A->B cable to link two units. You might think, by extension, that the TI-84+ used USB OTG. The answer is kind of! The USB implementation on the TI-84+ and TI-84+SE was very weird, and the OS didn't support anything other than TIConnect. Eventually the TI-84+CE introduced a much more standard USB controller, although I think support for any OTG peripheral still has to be hacked on to the OS. TI has always been at the forefront of calculator networking, and it has always been very weird and rarely used.

This solves part of the problem: it is clear, when you connect two phones, which should supply power and which should handle enumeration. The A-device is, by default, in charge. There are problems where this interacts with common USB devices types, though. One of the most common uses of USB with phones is mass storage (and its evil twin MTP). USB mass storage has a very strong sense of host and device at a logical level; the host can browse the devices files. When connecting two smartphones, though, you might want to browse from either end. Another common problem case here is that of the printer, or at least it would be if printer USB host support was ever usable. If you plug a printer into a phone, you might want to browse the phone as mass storage on the printer. Or you might want to use conventional USB printing to print a document from the phone's interface. In fact you almost certainly want to do the latter, because even with Android's extremely half-assed print spooler it's probably a lot more usable than the file browser your printer vendor managed to offer on its 2" resistive touchscreen.

OTG adds Host Negotiation Protocol, or HNP, to help in this situation. HNP allows the devices on a USB OTG connection to swap roles. While the A-device will always be the host when first connected, HNP can reverse the logical roles on demand.

This all sounds great, so where does it fall apart? Well, the usual places. Android devices often went a little off the script with their OTG implementations. First, the specification did not require devices to be capable of powering the bus, and phones couldn't. Fortunately that seems to have been a pretty short lived problem, only common in the first couple of generations of OTG devices. This wasn't the only limitation of OTG implementations; I don't have a good sense of scale but I've seen multiple reports that many OTG devices in the wild didn't actually support HNP, they just determined a role when connected based on the ID pin and could not change after that point.

Finally, and more insidiously, the whole thing about OTG devices having an AB connector didn't go over as well as intended. We actually must admire TI for their rare dedication to standards compliance. A lot of Android phones with OTG support had a micro-B connector only, and as a result a lot of OTG adapters use a micro-B connector.

There's a reason this was common; since A and B plugs are electrically differentiable regardless of the shape of the shell, the shell shape arguably doesn't matter. You could be a heavy OTG user with such a noncompliant phone and adapter and never notice. The problem only emerges when you get a (rare) standards-compliant OTG adapter or, probably more common, OTG A-B cable. Despite being electrically compatible, the connector won't fit into your phone. Of course this behavior feeds itself; as soon as devices with an improper B port were common, manufacturers of cables were greatly discouraged from using the correct A connector.

The downside, conceptually, is that you could plug an OTG A connector (with a B-shaped shell) into a device with no OTG support. In theory this could cause problems, in practice the problems don't seem to have been common since both devices would think they were B devices and (if standards compliant) not provide power. Essentially these improper OTG adapters create a B-B cable. It's a similar problem to an A-A cable but, in practice, less severe. Like an extension cord with two female ends. Home Depot might even help you make one of those.

While trying to figure out which iPhones had OTG support, I ran across an Apple Community thread where someone helpfully replied "I haven't heard of OTG in over a decade." Well, it's not a very helpful reply, but it's not exactly wrong either. No doubt the dearth of information on iOS OTG is in part because no one ever really cared. Much like the HDMI-over-USB support that a generation of Android phones included, OTG was an obscure feature. I'm not sure I have ever, even once, seen a human being other than myself make use of OTG.

Besides, it was completely buried by USB-C.

The thing is that OTG is not gone at all, in fact, it's probably more popular than ever before. There seems to be some confusion about how OTG has evolved with USB specifications. I came across more than one article saying that USB 3.1 Dual Role replaced OTG. This assertion is... confusing. It's not incorrect, but there's a good chance of it leading you int he wrong direction.

Much of the confusion comes from the fact that Dual-Role doesn't mean anything that specific. The term Dual-Role and various resulting acronyms like DRD and DRP have been applied to multiple concepts over the life of USB. Some vendors say "static dual role" to refer to devices that can be configured as either host or device (like the N770). Some vendors use dual role to identify chipsets that detect role based on the ID pin but are not actually capable of OTG protocols like HNP. Some articles use dual role to identify chipsets with OTG support. Subjectively, I think the intent of the changes in USB 3.1 were mostly to formally adopt the "dual role" term that was already the norm in informal use---and hopefully standardize the meaning.

For USB-C connectors, it's more complicated. USB-C cables are symmetric, they do not identify a host or device end in any way. Instead, the USB-C ports use resistance values to indicate their type. When either end indicates that it is only capable of the device role, the situation is simple, behaving basically the same way that OTG did: the host detects that the other end is a device and behaves as the host.

When both ends support the host role, things work differently: the Dual Role feature of USB-C comes into play. The actual implementation is reasonably simple; a dual-role USB-C controller will attempt to set up a connection both ways and go with whichever succeeds. There are some minor complications on top of this, for example, the controller can be configured with a "preference" for host or device role. This means that when you plug your phone into your computer via USB-C, the computer will assume the host role, because although it's capable of either the phone is configured with a preference for the device role. That matches consumer expectations. When both devices are capable of dual roles and neither specifies a preference, the outcome is random. This scenario is interesting but not all that common in practice.

The detection of host or device role by USB-C is based on the CC pins, basically a more flexible version of OTG's ID pin. There's another important difference between the behavior of USB-C and A/B: USB-C interfaces provide no power until they detect, via the CC pins, that the other device expects it. This is an important ingredient to mitigate the problem with A-A cables, that both devices will attempt to power the same bus.

The USB-C approach of using CC pins and having dual role controllers attempt one or the other at their preference is, for the most part, a much more elegant approach. There are a couple of oddities. First, in practice cables from C to A or B connectors are extremely common. These cables must provide the appropriate values on the CC pins to allow the USB-C controller to correctly determine its role, both for data and power delivery.

Second, what about role reversal? For type A and B connectors, this is achieved via HNP, but HNP is not supported on USB-C. Application notes from several USB controller vendors explain that, oddly enough, the only way to perform role reversal with USB-C is to implement USB Power Delivery (PD) and use the PD negotiation protocol to change the source of power. In other words, while OTG allows reversing host and device roles independently of the bus power source, USB-C does not. The end supplying power is always the host end. This apparent limitation probably isn't that big of a deal, considering that the role reversal feature of OTG was reportedly seldom implemented.

That's a bit of a look into what happens when you plug two USB hosts into each other. Are you confused? Yeah, I'm a little confused too. The details vary, and a lot more based on the capabilities of the individual devices rather than the USB version in use. This has been the malaise of USB for a solid decade now, at least: the specification has become so expansive, with so many non-mandatory features, that it's a crapshoot what capabilities any given USB port actually has. The fact that USB-C supports a bevy of alternate modes like Thunderbolt and HDMI only adds further confusion.

I sort of miss when the problem was just inappropriate micro-B connectors. Nonetheless, USB-C dual role support seems ubiquitous in modern smartphones, and that's the only place any of this ever really mattered. Most embedded devices still seem to prefer to just provide two USB ports: a host port and a device port. And no one ever uses the USB host support on their printer. It's absurd, no one ever would. Have you seen what HP thinks is a decent file browser? Good lord.

[1] My first smartphone was the HTC Thunderbolt. No one, not even me, will speak of that thing with nostalgia. It was pretty cool owning one of the first LTE devices on the market, though. There was no contention at all in areas with LTE service and I was getting 75+Mbps mobile tethering in 2011. Then everyone else had LTE too and the good times ended.

[2] There are actually several additional states defined by fixed resistances that tell the controller that it is the A-device but power will be supplied by the bus. These states were intended for Y-cables that allowed you to charge your phone from an external charger while using OTG. In this case neither device supplies power, the external charger does. The details of how this works are quite straightforward but will be confusing to keep adding as an exception, so I'm going to pretend the whole feature doesn't exist.

>>> 2023-12-23 ITT Technical Institute

Programming note/shameless plug: I am finally on Mastodon.

The history of the telephone industry is a bit of an odd one. For the greatest part of the 20th century, telephony in the United States was largely a monopoly of AT&T and its many affiliates. This wasn't always the case, though. AT&T held patents on their telephone implementation, but Bell's invention was not the only way to construct a practical telephone. During the late 19th century, telephone companies proliferated, most using variations on the design they felt would fall outside of Ma Bell's patent portfolio. AT&T was aggressive in challenging these operations but not always successful. During this period, it was not at all unusual for a city to have multiple competing telephone companies that were not interconnected.

Shortly after the turn of the 20th century, AT&T moved more decisively towards monopoly. Theodore Newton Vail, president of AT&T during this period, adopted the term "Universal Service" to describe the targeted monopoly state: there would be one universal telephone system. One operated under the policies and, by implication, the ownership of AT&T. AT&T's path to monopoly involved many political and business maneuvers, the details of which have filled more than a few dissertations in history and economics. By the 1920s the deal was done, there would be virtually no (and in a legal sense literally no) long-distance telephone infrastructure in the United States outside of The Bell System.

But what of the era's many telephone entrepreneurs? For several American telephone companies struggling to stand up to AT&T, the best opportunities were overseas. A number of countries, especially elsewhere in the Americas, had telephone systems built by AT&T's domestic competitors. Perhaps the most neatly named was ITT, the International Telephone and Telegraph company. ITT was formed from the combination of Puerto Rican and Cuban telephone companies, and through a series of acquisitions expanded into Europe.

Telefónica, for example, is a descendent of an early ITT acquisition. Other European acquisitions led to wartime complications, like the C. Lorenz company, which under ITT ownership functioned as a defense contractor to the Nazis during WWII. Domestically, ITT also expanded into a number of businesses outside of the monopolized telephone industry, including telegraphy and international cables.

ITT had been bolstered as well by an effect of AT&T's first round of antitrust cases during the 1910s and 1920s. As part of one of several settlements, AT&T agreed to divest several overseas operations to focus instead on the domestic market. They found a perfect buyer: ITT, a company which already seemed like a sibling of AT&T and through acquisitions came to function as one.

ITT grew rapidly during the mid-century, and in the pattern of many industrial conglomerates of the time ITT diversified. Brands like Sheraton Hotels and Avis Rent-a-Car joined the ITT portfolio (incidentally, Avis would be spun off, conglomerated with others, and then purchased by previous CAB subject Beatrice Foods). ITT was a multi-billion-dollar American giant.

Elsewhere in the early technology industry, salesman Howard W. Sams worked for the P. R. Mallory Company in Indianapolis during the 1930s and 1940s. Mallory made batteries and electronic components, especially for the expanding radio industry, and as Sams sold radio components to Mallory customers he saw a common problem and a sales opportunity: radio technicians often needed replacement components, but had a hard time identifying them and finding a manufacturer. Under the auspices of the Mallory company Sams produced and published several books on radio repair and electronic components, but Mallory didn't see the potential that Sams did in these technical manuals.

Sams, driven by the same electronics industry fervor as so many telephone entrepreneurs, struck out on his own. Incorporated in 1946, the Howard W. Sams Company found quick success with its Photofact series. Sort of the radio equivalent of Haynes and Chilton in the auto industry, Photofact provided schematics, parts lists, and repair instructions for popular radio receivers. They were often found on the shelves of both technicians and hobbyists, and propelled the Sams Company to million-dollar revenues by the early 1950s.

Sams would expand along with the electronics industry, publishing manuals on all types of consumer electronics and, by the 1960s, books on the use of computers. Sams, as a technical press, eventually made its way into the ownership of Pearson. Through Pearson's InformIT, the Sams Teach Yourself series remains in bookstores today. I am not quite sure, but I think one of the first technical books I ever picked up was an earlier edition of Sams HTML in 24 Hours.

The 1960s were an ambitious era, and Sams was not content with just books. Sams had taught thousands electronics technicians through their books. Many radio technicians had demonstrated their qualifications and kept up to date by maintaining a membership in the Howard Sams Radio Institute, a sort of correspondence program. It was a natural extension to teach electronics skills in person. In 1963, Sams opened the Sams Technical Institute in Indianapolis. Shortly after, they purchased the Acme Institute of Technology (Dayton, Ohio) and the charmingly named Teletronic Technical Institute (Evansville, Indiana), rebranding both as Sams campuses.

In 1965, the Sams Technical Institute had 2,300 students across five locations. Sams added the Bramwell Business College to its training division, signaling a move into the broader world of higher education. It was a fast growing business; it must have looked like a great opportunity to a telephone company looking for more ways to diversify. In 1968, ITT purchased the entire training division from Sams, renaming it ITT Educational Services [1].

ITT approached education with the same zeal it had overseas telephone service. ITT Educational Services spent the late '60s and early '70s on a shopping spree, adding campus after campus to the ITT system. Two newly constructed campuses expanded ITT's business programs, and during the '70s ITT introduced formal curriculum standardization programs and a bureaucratic structure to support its many locations. Along with expansion came a punchier name: the ITT Technical Institute.

"Tri-State Businessmen Look to ITT Business Institute, Inc. for Graduates," reads one corner of a 1970 full-page newspaper ad. "ITT adds motorcycle repair course to program," 1973. "THE ELECTRONICS AGE IS HERE. If your eyes are on the future, ITT Technical institute can prepare you for a HIGH PAYING, EXCITING career in... ELECTRONICS," 1971. ITT Tech has always known the value of advertising, and ran everything from full-page "advertorials" to succinct classified ads throughout their growing region.

During this period, ITT Tech clearly operated as a vocational school rather than a higher education institution. Many of its programs ran as short as two months, and they were consistently advertised as direct preparation for a career. These sorts of job-oriented programs were very attractive to veterans returning from Vietnam, and ITT widely advertised to veterans on the basis of its approval (clearly by 1972 based on newspaper advertisements, although some sources say 1974) for payment under the GI Bill. Around the same time ITT Tech was approved for the fairly new federal student loan program. Many of ITT's students attended on government money, with or without the expectation of repayment.

ITT Tech flourished. By the mid-'70s the locations were difficult to count, and ITT had over 1,000 students in several states. ITT Tech was the "coding boot camp" of its day, advertising computer programming courses that were sure to lead to employment in just about six months. Like the coding boot camps of our day, these claims were suspect.

In 1975, ITT Tech was the subject of investigations in at least two states. In Indiana, three students complained to the Evansville municipal government after ITT recruiters promised them financial aid and federally subsidized employment during their program. ITT and federal work study, they were told, would take care of all their living expenses. Instead, they ended up living in a YWCA off of food stamps. The Indiana board overseeing private schools allowed ITT to keep its accreditation only after ITT promised to rework its entire recruiting policy---and pointed out that the recruiters involved had left the company. ITT refunded the tuition of a dozen students who joined the complaint, which no doubt helped their case with the state.

Meanwhile, in Massachusetts, the Boston Globe ran a ten-part investigative series on the growing for-profit vocational education industry. ITT Tech, they alleged, promised recruits to its medical assistant program guaranteed post-graduation employment. The Globe claimed that almost no students of the program successfully found jobs, and the Massachusetts Attorney General agreed. In fact, the AG found, the program's placement rate didn't quite reach 5%. For a settlement, ITT Tech agreed to change its recruiting practices and refund nearly half a million dollars in tuition and fees.

ITT continued to expand at a brisk pace, adding more than a dozen locations in the early '80s and beginning to offer associates degrees. Newspapers from Florida to California ran ads exhorting readers to "Make the right connections! Call ITT Technical Institute." As the 1990s dawned, ITT Tech enjoyed the same energy as the computer industry, and aspired to the same scale. In 1992, ITT Tech announced their "Vision 2000" master plan, calling for bachelor's programs, 80 locations, and 45,000 students for beginning of the new millennium. ITT Tech was the largest provider of vocational training the country.

In 1993, ITT Tech was one of few schools accepted into the first year of the Direct Student Loan program. The availability of these new loans gave enrollment another boost, as ITT Tech reached 54 locations and 20,000 students. In 1994, ITT Tech started to gain independence from its former parent: an IPO sold 17% ownership to the open market, with ITT retaining the remaining 83%. The next year, ITT itself went through a reorganization and split, with its majority share of ITT Tech landing in the new ITT Corporation.

As was the case with so many diversified conglomerates of the '90s (see Beatrice Foods again), ITT's reorganization was a bad portent. ITT Hartford, the spun-out financial services division, survives today as The Hartford. ITT Industries, the spun-out defense contracting division, survives today as well, confusingly renamed to ITT Corporation. But the third part of the 1995 breakup, the ITT Corporation itself, merged with Starwood Hotels and Resorts. The real estate and hospitality side-business of a telephone and telegraph company saw the end of its parent.

Starwood had little interest in vocational education, and over the remainder of the '90s sold off its entire share of ITT Tech. Divestment was a good idea: the end of the '90s hit hard for ITT Tech. Besides the general decline of the tech industry as the dot com bubble burst, ITT Tech's suspect recruiting practices were back. This time, they had attracted federal attention.

In 1999, two ITT Tech employees filed a federal whistleblower suit alleging that ITT Tech trained recruiters to use high-pressure sales tactics and outright deception to obtain students eligible for federal aid. Recruiters were paid a commission for each student they brought in, and ITT Tech obtained 70% of its revenue from federal aid programs. A federal investigation moved slowly, apparently protracted by the Department of Education's nervous approach following the criticism it received for shutting down similar operation Computer Learning Centers. In 2004, federal agents raided ITT Tech campuses across ten states, collecting records on recruitment and federal funding.

During the early 2000s ITT Tech students defaulted on $400 million in federal student loans. The result, a large portion of ITT Tech revenue coming from defaulted federal loans, attracted ongoing attention. ITT Tech was deft in its legal defense, though, and through a series of legal victories and, more often, settlements, ITT Tech stayed in business.

ITT Tech aggressively advertised throughout its history. In the late '90s and early '00s, ITT Tech's constant television spots filled a corner of my brain. "How Much You Know Measures How Far You Can Go," a TV spot proclaims, before ITT's distinctive block letter logo faded on screen in metallic silver. By the year 2000, International Telephone and Telegraph, or rather its scattered remains, no longer had any relationship with ITT Tech. Starwood agreed to license the name and logo to the independent public ITT Technical Institutes corporation, though, and with the decline of ITT's original business the ITT name and logo became associated far more with the for-profit college than the electronics manufacturer.

For-profit universities attracted a lot of press in the '00s---the wrong kind of press. ITT Tech was far from unique in suspicious advertising and recruiting, high tuition rates, and frequent defaults on the federal loans that covered that tuition. For-profit education, it seemed, was more of a scam on the taxpayer dollar than way to secure a promising new career. Publicly traded colleges like DeVry and the University of Phoenix had repeated scandals over their use, or abuse, of federal aid, and a 2004 criminal investigation into ITT Tech for fraud on federal student aid made its future murky.

ITT Tech was a survivor. The criminal case fell apart, the whistleblower lawsuit lead to nothing, and ITT Tech continued to grow. In 2009, ITT Tech acquired the formerly nonprofit Daniel Webster University, part of a wave of for-profit conversions of small colleges. ITT Tech explained the purchase as a way to expand their aeronautics offerings, but observers suspected other motives, ones that had more to do with the perceived legitimacy of what was once a nonprofit, regionally accredited institution. Today, regional accreditors re-investigate institutions that are purchased. There was a series of suspect expansions of small colleges to encompass large for-profit organizations during the '00s that lead to the tightening of these rules.

ITT Tech, numerically, achieved an incredible high. In 2014, ITT Tech reported a total cost of attendance of up to $85,000. I didn't spend that much on my BS and MS combined. Of course, I attended college in impoverished New Mexico, but we can make a comparison locally. ITT Tech operated here as well, and curiously, New Mexico tuition is specially listed in an ITT Tech cost estimate report because it is higher. At its location in Albuquerque's Journal Center office development, ITT Tech charged more than $51,000 in tuition alone for an Associate's in Criminal Justice. The same program at Central New Mexico Community College would have cost under $4,000 over the two years [2].

That isn't the most remarkable, though. A Bachelor's in Criminal Justice would run over $100,000---more than the cost of a JD at UNM School of Law, for an out-of-state student, today.

In 2014, more than 80% of ITT Tech's revenue came from federal student aid. Their loan default rate was the highest of even for-profit programs. With their extreme tuition costs and notoriously poor job placement rates, ITT Tech increasingly had the appearance of an outright fraud.

Death came swiftly for ITT Tech. In 2016, they were a giant with more than 130 campuses and 40,000 students. The Consumer Financial Protection Bureau sued. State Attorneys General followed, with New Mexico's Hector Balderas one of the first two. The killing blow, though, came from the Department of Education, which revoked ITT Tech's eligibility for federal student aid. Weeks later, ITT Tech stopped accepting applications. The next month, they filed for bankruptcy, chapter 7, liquidation.

Over the following years, the ITT Tech scandal would continue to echo. After a series of lawsuits, the Department of Education agreed to forgive the federal debt of ITT Tech attendees, although a decision by Betsy DeVos to end the ITT Tech forgiveness program produced a new round of lawsuits over the matter in 2018. Private lenders faced similar lawsuits, and made similar settlements. Between federal and private lenders, I estimate almost $4.5 billion in loans to pay ITT Tech tuition were written off.

The Department of Education decision to end federal aid to ITT Tech was based, in part, on ITT Tech's fraying relationship with its accreditor. The Accrediting Council for Independent Colleges and Schools (ACICS), a favorite of for-profit colleges, had its own problems. That same summer in 2016, the Department of Education ended federal recognition of ACICS. ACICS accreditation reviews had been cursory, and it routinely continued to accredit colleges despite their failure to meet even ACIC's lax standards. ITT Tech was not the only large ACIC-accredited institution to collapse in scandal.

Two years later, Betsy DeVos reinstated ACICS to federal recognition. Only 85 institutions still relied on ACICS, such august names as the Professional Golfers Career College and certain campuses of the Art Institutes that were suspect even by the norms of the Art Institutes (the Art Institutes folded just a few months ago following a similar federal loan fraud scandal). ACICS lost federal recognition again in 2022. Only time will tell what the next presidential administration holds for the for-profit college industry.

ITT endured a long fall from grace. A leading electronics manufacturer in 1929, a diversified conglomerate in 1960, scandals through the 1970s. You might say that ITT is distinctly American in all the best and worst ways. They grew to billions in revenue through an aggressive program of acquisitions. They were implicated in the CIA coup in Chile. They made telephones and radios and radars and all the things that formed the backbone of the mid-century American electronics industry.

The modern ITT Corporation, descended from spinoff company ITT Industries, continues on as an industrial automation company. They have abandoned the former ITT logo, distancing themselves from their origin. The former defense division became Exelis, later part of Harris, now part of L3, doomed to slowly sink into the monopolized, lethargic American defense industry. German tool and appliance company Kärcher apparently holds a license to the former ITT logo, although I struggle to find any use of it.

To most Americans, ITT is ITT Tech, a so-called college that was actually a scam, an infamous scandal, a sink of billions of dollars in federal money. Dozens of telephone companies around the world, tracing their history back to ITT, are probably better off distancing themselves from what was once a promising international telephone operator, a meaningful technical competitor to Western Electric. The conglomeration of the second half of the 20th century put companies together and then tore them apart; they seldom made it out in as good of condition as they went in. ITT went through the same cycle as so many other large American corporations. They went into hotels, car rentals, then into colleges. They left thousands of students in the lurch on the way out. When ITT Tech went bankrupt, everyone else had already started the semester. They weren't accepting applicants. They wouldn't accept transfer credit from ITT anyway; ITT's accreditation was suspect.

"What you don't know can hurt you," a 1990s ITT Tech advertisement declares. In Reddit threads, ITT Tech alums debate if they're better off telling prospective employers they never went to college at all.

[1] Sources actually vary on when ITT purchased Sams Training Institute, with some 1970s newspaper articles putting it as early as 1966, but 1968 is the year that ITT's involvement in Sams was advertised in the papers. Further confusing things, the former Sams locations continued to operate under the Sams Technical Institute name until around 1970, with verbiage like "part of ITT Educational Services" inconsistently appearing. ITT may have been weighing the value of its brand recognition against Sams but apparently made a solid decision during 1970, after which ads virtually always use the ITT name and logo above any other.

[2] Today, undergraduate education across all of New Mexico's public universities and community colleges is free for state residents. Unfortunately 2014 was not such an enlightened time. I must take every opportunity to brag about this remarkable and unusual achievement in our state politics.

>>> 2023-12-05 vhf omnidirectional range

The term "VHF omnidirectional range" can at first be confusing, because it includes "range"---a measurement that the technology does not provide. The answer to this conundrum is, as is so often the case, history. The "range" refers not to the radio equipment but to the space around it, the area in which the signal can be received. VOR is an inherently spatial technology; the signal is useless except as it relates to the physical world around it.

This use of the word "range" is about as old as instrument flying, dating back to the first radionavigation devices in the 1930s. We still use it today, in the somewhat abstract sense of an acronym that is rarely expanded: VOR.

This is Truth or Consequences VOR. Or, perhaps more accurately, the transmitter that defines the center of the Truth or Consequences VOR, which extends perhaps two hundred miles around this point. The range can be observed only by instruments, but it's there, a phase shift that varies like terrain.

The basic concept of VOR is reasonably simple: a signal is transmitted with two components, a 30Hz tone in amplitude modulation and a 30Hz in frequency modulation. The two tones are out of phase, by an amount that is determined by your position in the range, and more specifically by the radial from the VOR transmitter to your position. This apparent feat of magic, a radio signal that is different in different locations, is often described as "space modulation."

The first VOR transmitters achieved this effect the obvious way, by rapidly spinning a directional antenna in time with the electronically generated phase shift. Spinning anything quickly becomes a maintenance headache, and so VOR was quickly transitioned to solid-state techniques. Modern VOR transmitters are electronically rotated, by one of two techniques. They rotate in the same sense as images on a screen, a set of discrete changes in a solid state system that produce the effect of rotation.

The Truth or Consequences VOR operates on 112.7 MHz, near the middle of the band assigned for this use. Patterned after the nearby Truth or Consequences Airport, KTCS, it identifies itself by transmitting "TCS" in Morse code. Modern charts give this identifier in dots and dashes, an affordance to the poor level of Morse literacy among contemporary pilots.

In the airspace, it defines the intersection of several airways. They all go generally north-south, unsurprising considering that the restricted airspace of White Sands Missile Range prevents nearly all flight to the east. Flights following the Rio Grande, most north-south traffic in this area, will pass directly overhead on their way to VOR transmitters at Socorro or Deming or El Paso, where complicated airspace leads to two such sites very nearby.

This is the function that VORs serve: for the most part, you fly to or from them. Because the radial from the VOR to you remains constant, they provide a reliable and easy to use indication that you are still on the right track. A warning sign, verbose by tradition, articulates the significance:

This facility is used in FAA air traffic control. Loss of human life may result from service interruption. Any person who interferes with air traffic control or damages or trespasses on this property will be prosecuted under federal law.

The sign is backed up by a rustic wooden fence. Like most VOR transmitters, this one was built in the late 1950s or 1960s. The structure has seen only minimal changes since then, although the radio equipment has been improved and simplified.

The central, omnidirectional antenna of a VOR transmitter makes for a distinctive silhouette. You have likely noticed one before. I must admit that I have somewhat simplified; most of the volume of the central antenna housing is actually occupied by the TACAN antenna. Most VOR sites in the US are really VORTAC sites, combining the civilian VOR and military TACAN systems into one facility. TACAN has several minor advantages over VOR for military use, but one big advantage: it provides not only a radial but a distance. The same system used by TACAN for distance information, based on an unusual radio modulation technique called "squitter," can be used by civilian aircraft as well in the form of DME. VORTAC sites thus provide VOR, DME, and TACAN service.

True VOR sites, rare in the US but plentiful across the rest of the world, have smaller central antennas. If you are not used to observing the ring of radial antennas, you might not recognize them as the same system.

The radial antennas are placed in a circle some distance away, to open space between them. This reduces, but does not eliminate, the effect of each antenna's radiated power being absorbed by its neighbors. They are often on the roof of the equipment building, and may be surrounded by a metallic ground plane that extends still further. Most US VORTAC sites, originally built before modern RF technology, rely on careful positioning on suitable terrain rather than a ground plane.

Intriguingly, the radial antennas are not directional designs. In a modern VOR site, the radial antennas transmit an in-phase signal. The phase shift used for space modulation is created by rapidly changing the omnidirectional antenna in use. The space modulation is created not by rotating the antenna, but by moving the antenna through a circular path and allowing the Doppler effect to vary the apparent phase of the received signal.

The lower part of the central antenna, the more cone shaped part, is mostly empty. It encloses the structure that supports the cylindrical radome that houses the actual antenna elements. In newer installations it is often an exposed frame, but the original midcentury sites all provide a conical enclosure. I suspect the circular metallic sheathing simplified calculation of the effective radiation pattern at the time.

An access door can be used to reach the interior to service the antennas; the rope holding this one closed is not standard equipment but is perhaps also not very unusual. These are old facilities. When this cone was installed, adjacent Interstate 25 wasn't an interstate yet.

Aviation engineers leave little to chance, and almost never leave a system without a spare. Ground-based infrastructure is no exception. Each VOR transmitter is continuously tested by a monitoring system. A pair of antennas mounted on a post near the fence line feed redundant monitoring systems that ensure the static antennas receive the correct radial. If failure or a bad fix are detected, it switches the transmit antennas over to a second, redundant set of radio equipment. The problem is reported to the FAA, and Tech Ops staff are dispatched to investigate the problem.

Occasionally, the telephone lines VOR stations use to report problems are, themselves, unreliable. When Tech Ops is unable to remotely monitor a VOR station, they issue a NOTAM that it should not be relied upon.

The rear of the building better shows its age. The wall is scarred where old electrical service equipment has been removed; the weather-tight light fixture is a piece of incandescent history. It has probably been broken for longer than I have been alive.

A 1000 gallon propane tank to one side will supply the generator in the enclosure in case of a failure. Records of the Petroleum Storage Bureau of the New Mexico Environment Department show that an underground fuel tank was present at this site but has been removed. Propane is often selected for newer standby generator installations where an underground tank, no longer up to environmental safety standards, had to be removed.

It is indeed in its twilight years. The FAA has shut down about half of the VOR transmitters. TCS was spared this round, with all but one of the VOR transmitters in sparsely covered New Mexico. It is part of the "minimum operational network." It remains to be seen how long VOR's skeleton crew will carry on. A number of countries have now announced the end of VOR service. Another casualty to satellite PNT, joining LORAN wherever dead radio systems go.

The vastness and sparse population of southern New Mexico pose many challenges. One the FAA has long had to contend with is communications. Very near the Truth or Consequences VOR transmitter is an FAA microwave relay site. This tower is part of a chain that relays radar data from southern New Mexico to the air route traffic control center in Albuquerque.

When it was first built, the design of microwave communications equipment was much less advanced than it is today. Practical antennas were bulky and often pressurized for water tightness. Waveguides were expensive and cables were inefficient. To ease maintenance, shorten feedlines, and reduce tower loading, the actual antennas were installed on shelves near the bottom of the tower, pointing straight upwards. At the top of the tower, two passive reflectors acted like mirrors to redirect the signal into the distance. This "periscope" design was widely used by Western Union in the early days of microwave data networking.

Today, this system is partially retired, replaced by commercial fiber networks. This tower survives, maintained under contract by L3Harris. As the compound name suggests, half of this company used to Harris, a pioneer in microwave technology. The other half used to be L3, which split off from Lockheed Martin, which bought it when it was called Loral. Loral was a broad defense contractor, but had its history and focus in radar, another application of microwave RF engineering.

Two old radio sites, the remains of ambitious nationwide systems that helped create today's ubiquitous aviation. A town named after an old radio show. Some of the great achievements of radio history are out there in Sierra County.

>>> 2023-11-25 the curse of docker

I'm heading to Las Vegas for re:invent soon, perhaps the most boring type of industry extravaganza there could be. In that spirit, I thought I would write something quick and oddly professional: I'm going to complain about Docker.

Packaging software is one of those fundamental problems in system administration. It's so important, so influential on the way a system is used, that package managers are often the main identity of operating systems. Consider Windows: the operating system's most alarming defect in the eyes of many "Linux people" is its lack of package management, despite Microsoft's numerous attempts to introduce the concept. Well, perhaps more likely, because of the number of those attempts. And still, in the Linux world, distributions are differentiated primarily by their approach to managing software repositories. I don't just mean the difference between dpkg and rpm, but rather more fundamental decisions, like opinionated vs. upstream configuration and stable repositories vs. a rolling release. RHEL and Arch share the vast majority of their implementation and yet have very different vibes.

Linux distributions have, for the most part, consolidated on a certain philosophy of how software ought to be packaged, if not how often. One of the basic concepts shared by most Linux systems is centralization of dependencies. Libraries should be declared as dependencies, and the packages depended on should be installed in a common location for use of the linker. This can create a challenge: different pieces of software might depend on different versions of a library, which may not be compatible. This is the central challenge of maintaining a Linux distribution, in the classical sense: providing repositories of software versions that will all work correctly together. One of the advantages of stable distributions like RHEL is that they are very reliable in doing this; one of the disadvantages is that they achieve that goal by packaging new versions very infrequently.

Because of the need to provide mutually compatible versions of a huge range of software, and to ensure compliance with all kinds of other norms established by distributions (which may range from philosophical policies like free software to rules on the layout of configuration files), putting new software into Linux distributions can be... painful. For software maintainers, it means dealing with a bunch of distributions using a bunch of old versions with various specific build and configuration quirks. For distribution and package maintainers, it means bending all kinds of upstream software into compliance with distribution policy and figuring out version and dependency problems. It's all a lot of work, and while there are some norms, in practice it's sort of a wild scramble to do the work to make all this happen. Software developers that want their software to be widely used have to put up with distros. Distros that want software have to put up with software developers. Everyone gets mad.

Naturally there have been various attempts to ease these problems. Naturally they are indeed various and the community has not really consolidated on any one approach. In the desktop environment, Flatpak, Snap, and AppImage are all distressingly common ways of distributing software. The images or applications for these systems package the software complete with its dependencies, providing a complete self-contained environment that should work correctly on any distribution. The fact that I have multiple times had to unpack flatpaks and modify them to fix dependencies reveals that this concept doesn't always work entirely as advertised, but to be fair that kind of situation usually crops up when the software has to interact with elements of the system that the runtime can't properly isolate them from. The video stack is a classic example, where errant OpenGL libraries in packages might have to be removed or replaced for them to function with your particular graphics driver.

Still, these systems work reasonably well, well enough that they continue to proliferate. They are greatly aided by the nature of the desktop applications for which they're used (Snapcraft's system ambitions notwithstanding). Desktop applications tend to interact mostly with the user and receive their configuration via their own interface. Limiting the interaction surface mostly to a GUI window is actually tremendously helpful in making sandboxing feasible, although it continues to show rough edges when interacting with the file system.

I will note that I'm barely mentioning sandboxing here because I'm just not discussing it at the moment. Sandboxing is useful for security and even stability purposes, but I'm looking at these tools primarily as a way of packaging software for distribution. Sandboxed software can be distributed by more conventional means as well, and a few crusty old packages show that it's not as modern of a concept as it's often made out to be.

Anyway, what I really wanted to complain a bit about is the realm of software intended to be run on servers. Here, there is a clear champion: Docker, and to a lesser degree the ecosystem of compatible tools like Podman. The release of Docker lead to a surprisingly rapid change in what are widely considered best practices for server operations. While Docker images a means of distributing software first seemed to appeal mostly to large scalable environments with container orchestration, it sort of merged together with ideas from Vagrant and others to become a common means of distributing software for developer and single-node use as well.

Today, Docker is the most widespread way that server-side software is distributed for Linux. I hate it.

This is not a criticism of containers in general. Containerization is a wonderful thing with many advantages, even if the advantages over lightweight VMs are perhaps not as great as commonly claimed. I'm not sure that Docker has saved me more hours than it's cost, but to be fair I work as a DevOps consultant and, as a general rule, people don't get me involved unless the current situation isn't working properly. Docker images that run correctly with minimal effort don't make for many billable hours.

What really irritates me these days is not really the use of Docker images in DevOps environments that are, to some extent, centrally planned and managed. The problem is the use of Docker as a lowest common denominator, or perhaps more accurately lowest common effort, approach to distributing software to end users. When I see open-source, server-side software offered to me as a Docker image or--even worse---Docker Compose stack, my gut reaction is irritation. These sorts of things usually take longer to get working than equivalent software distributed as a conventional Linux package or to be built from source.

But wait, how does that happen? Isn't Docker supposed to make everything completely self-contained? Let's consider the common problems, something that I will call my Taxonomy of Docker Gone Bad.

Configuration

One of the biggest problems with Docker-as-distribution is the lack of consistent conventions for configuration. The vast majority of server-side Linux software accepts its configuration through an ages-old technique of reading a text file. This certainly isn't perfect! But, it is pretty consistent in its general contours. Docker images, on the other hand...

If you subscribe to the principles of the 12-factor-app, the best way for a Docker image to take configuration is probably via environment variables. This has the upside that it's quite straightforward to provide them on the command line when starting the container. It has the downside that environment variables aren't great for conveying structured data, and you usually interact with them via shell scripts that have clumsy handling of long or complicated values. A lot of Docker images used in DevOps environments take their configuration from environment variables, but they tend to make it a lot more feasible by avoiding complex configuration (by assuming TLS will be terminated by "someone else" for example) or getting a lot of their configuration from a database or service on the network.

For most end-user software though, configuration is too complex or verbose to be comfortable in environment variables. So, often, they fall back to configuration files. You have to get the configuration file into the container's file system somehow, and Docker provides numerous ways of doing so. Documentation on different packages will vary on which way it recommends. There are frequently caveats around ownership and permissions.

Making things worse, a lot of Docker images try to make configuration less painful by providing some sort of entry-point shell script that generates the full configuration from some simpler document provided to the container. Of course this level of abstraction, often poorly documented or entirely undocumented in practice, serves mostly to make troubleshooting a lot more difficult. How many times have we all experienced the joy of software failing to start, referencing some configuration key that isn't in what we provided, leading us to have to find have the Docker image build materials and read the entrypoint script to figure out how it generates that value?

The situation with configuration entrypoint scripts becomes particularly acute when those scripts are opinionated, and opinionated is often a nice way of saying "unsuitable for any configuration other than the developer's." Probably at least a dozen times I have had to build my own version of a Docker image to replace or augment an entrypoint script that doesn't expose parameters that the underlying software accepts.

In the worst case, some Docker images provide no documentation at all, and you have to shell into them and poke around to figure out where the actual configuration file used by the running software is even located. Docker images must always provide at least some basic README information on how the packaged software is configured.

Filesystems

One of the advantages of Docker is sandboxing or isolation, which of course means that Docker runs into the same problem that all sandboxes do. Sandbox isolation concepts do not interact well with Linux file systems. You don't even have to get into UID behavior to have problems here, just a Docker Compose stack that uses named volumes can be enough to drive you to drink. Everyday operations tasks like backups, to say nothing of troubleshooting, can get a lot more frustrating when you have to use a dummy container to interact with files in a named volume. The porcelain around named volumes has improved over time, but seemingly simple operations can still be weirdly inconsistent between Docker versions and, worse, other implementations like Podman.

But then, of course, there's the UID thing. One of the great sins of Docker is having normalized running software as root. Yes, Docker provides a degree of isolation, but from a perspective of defense in depth running anything with user exposure as root continues to be a poor practice. Of course this is one thing that often leads me to have to rebuild containers provided by software projects, and a number of common Docker practices don't make it easy. It all gets much more complicated if you use hostmounts because of UID mapping, and slightly complex environments with Docker can turn into NFS-style puzzles around UID allocation. Mitigating this mess is one of the advantages to named volumes, of course, with the pain points they bring.

Non-portable Containers

The irony of Docker for distribution, though, and especially Docker Compose, is that there are a lot of common practices that negatively impact portability---ostensibly the main benefit of this approach. Doing anything non-default with networks in Docker Compose will often create stacks that don't work correctly on machines with complex network setups. Too many Docker Compose stacks like to assume that default, well-known ports are available for listeners. They enable features of the underlying software without giving you a way to disable them, and assume common values that might not work in your environment.

One of the most common frustrations, for me personally, is TLS. As I have already alluded to, I preach a general principle that Docker containers should not terminate TLS. Accepting TLS connections means having access to the private key material. Even if 90-day ephemeral TLS certificates and a general atmosphere of laziness have deteriorated our discipline in this regard, private key material should be closely guarded. It should be stored in only one place and accessible to only one principal. You don't even have to get into these types of lofty security concerns, though. TLS is also sort of complicated to configure.

A lot of people who self-host software will have some type of SNI or virtual hosting situation. There may be wildcard certificates for multiple subdomains involved. All of this is best handled at a single point or a small number of dedicated points. It is absolutely maddening to encounter Docker images built with the assumption that they will individually handle TLS. Even with TLS completely aside, I would probably never expose a Docker container with some application directly to the internet. There are too many advantages to having a reverse proxy in front of it. And yet there are Docker Compose stacks out there for end-user software that want to use ACME to issue their own certificate! Now you have to dig through documentation to figure out how to disable that behavior.

The Single-Purpose Computer

All of these complaints are most common with what I would call hobby-tier software. Two examples that pop into my mind are HomeAssistant and Nextcloud. I don't call these hobby-tier to impugn the software, but rather to describe the average user.

Unfortunately, the kind of hobbyist that deploys software has had their mind addled by the cheap high of the Raspberry Pi. I'm being hyperbolic here, but this really is a problem. It's absurd the number of "self-hosted" software packages that assume they will run on dedicated hardware. Having "pi" in the name of a software product is a big red flag in my mind, it immediately makes me think "they will not have documented how to run this on a shared device." Call me old-fashioned, but I like my computers to perform more than one task, especially the ones that are running up my power bill 24/7.

HomeAssistant is probably the biggest offender here, because I run it in Docker on a machine with several other applications. It actively resists this, popping up an "unsupported software detected" maintenance notification after every update. Can you imagine if Postfix whined in its logs if it detected that it had neighbors?

Recently I decided to give NextCloud a try. This was long enough ago that the details elude me, but I think I burned around two hours trying to get the all-in-one Docker image to work in my environment. Finally I decided to give up and install it manually, to discover it was a plain old PHP application of the type I was regularly setting up in 2007. Is this a problem with kids these days? Do they not know how to fill in the config.php?

Hiding Sins

Of course, you will say, none of these problems would be widespread of people just made good Docker images. And yes, that is completely true! Perhaps one of the problems with Docker is that it's too easy to use. Creating an RPM or Debian package involves a certain barrier to entry, and it takes a whole lot of activation energy for even me to want to get rpmbuild going (advice: just use copr and rpkg). At the core of my complaints is the fact that distributing an application only as a Docker image is often evidence of a relatively immature project, or at least one without anyone who specializes in distribution. You have to expect a certain amount of friction in getting these sorts of things to work in a nonstandard environment.

It is a palpable irony, though, that Docker was once heralded as the ultimate solution to "works for me" and yet seems to just lead to the same situation existing at a higher level of configuration.

Last Thoughts

This is of course mostly my opinion and I'm sure you'll disagree on something, like my strong conviction that Docker Compose was one of the bigger mistakes of our era. Fifteen years ago I might have written a nearly identical article about all the problems I run into with RPMs created by small projects, but what surprises me about Docker is that it seems like projects can get to a large size, with substantial corporate backing, and still distribute in the form of a decidedly amateurish Docker Compose stack. Some of it is probably the lack of distribution engineering personnel on a lot of these projects, since Docker is "simple." Some of it is just the changing landscape of this class of software, with cheap single-board computers making Docker stacks just a little less specialized than a VM appliance image more palatable than they used to be. But some if it is also that I'm getting older and thus more cantankerous.

>>> 2023-11-19 Centrex

I have always been fascinated by the PABX - the private automatic branch exchange, often shortened to "PBX" in today's world where the "automatic" is implied. (Relatively) modern small and medium business PABXs of the type I like to collect are largely solid-state devices that mount on the wall. Picture a cabinet that's maybe two feet wide, a foot and half tall, and five inches deep. That's a pretty accurate depiction of my Comdial hybrid key/PABX system, recovered from the offices of a bankrupt publisher of Christian home schooling materials.

These types of PABX, now often associated with Panasonic on the small end, are affordable and don't require much maintenance or space. They have their limitations, though, particularly in terms of extension count. Besides, the fact that these compact PABX are available at all is the result of decades of development in electronics.

Not that long ago, PABX were far more complex. Early PBX systems were manual, and hotels were a common example of a business that would have a telephone operator on staff. The first PABX were based on the same basic technology as their contemporary phone switches, using step-by-step switches or even crossbar mechanisms. They no longer required an operator to connect every call, but were still mostly designed with the assumption that an attendant would handle some situations. Moreover, these early PABX were large, expensive, and required regular maintenance. They were often leased from the telephone company, and the rates weren't cheap.

PABX had another key limitation as well: they were specific to a location. Each extension had to be home-run wired to the PABX, easy in a single building but costly at the level of a campus and, especially, with buildings spread around a city. For organizations with distributed buildings like school districts, connecting extensions back to a central PABX could be significantly more expensive than connecting them to the public telephone exchange.

This problem must have been especially common in a city the size of New York, so it's no surprise that New York Telephone was the first to commercialize an alternative approach: Centrex.

Every technology writer must struggle with the temptation to call every managed service in history a precursor to "the Cloud." I am going to do my very best to resist that nagging desire, but it's difficult not to note the similarity between Centrex service and modern cloud PABX solutions. Indeed, Centrex relied on capabilities of telephone exchange equipment that are recognizably similar to mainframe computer concepts like LPARs and virtualization today. But we'll get there in a bit. First, we need to talk about what Centrex is.

I've had it in my mind to write something about Centrex for years, but I've always had a hard time knowing where to start. The facts about Centrex are often rather dry, and the details varied over years of development, making it hard to sum up the capabilities in short. So I hope that you will forgive this somewhat dry post. It covers something that I think is a very important part of telephone history, particularly from the perspective of the computer industry today. It also lists off a lot of boring details. I will try to illustrate with interesting examples everywhere I can. I am indebted, for many things but here especially, to many members of the Central Office mailing list. They filled in a lot of details that solidified my understanding of Centrex and its variants.

The basic promise of Centrex was this: instead of installing your own PABX, let the telephone company configure their own equipment to provide the features you want to your business phones. A Centrex line is a bit like a normal telephone line, but with all the added capabilities of a business phone system: intercom calling, transfers, attendants, routing and long distance policies, and so on. All of these features were provided by central telephone exchanges, but your lines were partitioned to be interconnected within your business.

Centrex was a huge success. By 1990, a huge range of large institutions had either started their telephone journey with Centrex or transitioned away from a conventional PABX and onto Centrex. It's very likely that you have interacted with a Centrex system before and perhaps not realized. And now, Centrex's days are numbered. Let's look at the details.

Centrex is often explained as a reuse of the existing central office equipment to serve PABX requirements. This isn't entirely incorrect, but it can be misleading. It was not all that unusual for Centrex to rely on equipment installed at the customer site, but operated by the telco. For this reason, it's better to think of Centrex as a managed service than as a "cloud" service, or a Service-as-a-Service, or whatever modern term you might be tempted to apply.

Centrex existed in two major variants: Centrex-CO and Centrex-CU. The CO case, for Central Office, entailed this well-known design of each business telephone line connecting to an existing telco central office, where a switch was configured to provide Centrex features on that line group. CU, for Customer Unit, looks more like a very large PABX. These systems were usually limited to very large customers, who would provide space for the telco to build a new central office on the customer's site. The exchange was located with the customer, but operated by the telco.

These two different categories of service lead to two different categories of customers, with different needs and usage patterns. Centrex-CO appealed to smaller organizations with fewer extensions, but also to larger organizations with extensions spread across a large area. In that case, wiring every extension back to the CO using telco infrastructure was less expensive than installing new wiring to a CU exchange. A prototypical example might be a municipal school district.

Centrex-CU appealed to customers with a large number of extensions grouped in a large building or a campus. In this case it was much less costly to wire extensions to the new CU site than to connect them all over the longer distance to an existing CO. A prototypical Centrex-CU customer might be a university.

Exactly how these systems worked varied greatly from exchange to exchange, but the basic concept is a form of partitioning. Telephone exchanges with support for Centrex service could be configured such that certain lines were grouped together and enabled for Centrex features. The individual lines needed to have access to Centrex-specific capabilities like service codes, but also needed to be properly associated with each other so that internal calling would indeed be internal to the customer. This concept of partitioning telephone switches had several different applications, and Western Electric and other manufacturers continued to enhance it until it reached a very high level of sophistication in digital switches.

Let's look at an example of a Centrex-CO. The State of New Mexico began a contract with Mountain States Telephone and Telegraph [1] for Centrex service in 1964. The new Centrex service replaced 11 manual switchboards distributed around Santa Fe, and included Wide-Area Telephone Service (WATS), a discount arrangement for long-distance calls placed from state offices to exchanges throughout New Mexico. On November 9th, 1964, technicians sent to Santa Fe by Western Electric completed the cutover at the state capitol complex. Incidentally, the capitol phones of the day were being installed in what is now the Bataan Memorial Building: construction of the Roundhouse, today New Mexico's distinctive state capitol, had just begun that same year.

The Centrex service was estimated to save $12,000 per month in the rental and operation of multiple state exchanges, and the combination of WATS and conference calling service was expected to produce further savings by reducing the need for state employees to travel for meetings. The new system was evidently a success, and lead to a series of minor improvements including a scheme later in 1964 to ensure that the designated official phone number of each state agency would be answered during the state lunch break (noon to 1:15). In 1965, Burns Reinier resigned her job as Chief Operator of the state Centrex to launch a campaign for Secretary of State. Many state employees would probably recognize her voice, but that apparently did not translate to recognition on the ballot, as she lost the Democratic party nomination to the Governor's former secretary.

The late 1960s saw a flurry of newspaper advertisements giving new phone numbers for state and municipal agencies, Albuquerque Public Schools, and universities, as they all consolidated onto the state-run Centrex system. Here we must consider the geographical nature of Centrex: Centrex service operates within a single telephone exchange. To span the gap between the capitol in Santa Fe, state offices and UNM in Albuquerque, NMSU in Las Cruces, and even the State Hospital in Las Vegas (NM), a system of tie lines were installed between Centrex facilities in each city. These tie lines were essentially dedicated long distance trunks leased by the state to connect calls between Centrex exchanges at lower cost than even WATS long-distance service.

This system was not entirely CO-based: in Albuquerque, a Centrex exchange was installed in state-leased space at what was then known as the National Building, 505 Marquette. In the late '60s, 505 Marquette also hosted Telepak, an early private network service from AT&T. It is perhaps a result of this legacy that 505 Marquette houses one of New Mexico's most important network facilities, a large carrier hotel now operated by H5 Data Centers. The installation of the Centrex exchange at 505 Marquette saved a lot of expense on new local loops, since a series of 1960s political and bureaucratic events lead to a concentration of state offices in the new building.

Having made this leap to customer unit systems, let's jump almost 30 years forward to an example of a Centrex-CU installation... one with a number of interesting details. In late 1989, Sandia National Laboratories ended its dependence on the Air Force for telephony services by contracting with AT&T for the installation of a 5ESS telephone exchange. The 5ESS, a digital switch and a rather new one at the time, brought with it not just advanced calling features but something even more compelling to an R&D institution at the time: data networking.

The Sandia installation went nearly all-in on ISDN, the integrated digital telephony and data standard that largely failed to achieve adoption for telephone applications. Besides the digital telephone sets, though, Sandia made full use of the data capabilities of the exchange. Computers connected to the data ports on the ISDN user terminals (the conventional term for the telephone instrument itself in an ISDN network) could make "data calls" over the telephone system to access IBM mainframes and other corporate computing resources... all at a blistering 64 kbps, the speed of an ISDN basic rate interface bearer channel. The ISDN network could even transport video calls, by combining multiple BRIs for 384 kbps aggregate capacity.

The 5ESS was installed on a building on Air Force property near Tech Area 1, and the 5ESS's robust support for remote switch modules was fully leveraged to place an RSM in each Tech Area. The new system required renumbering, always a hassle, but allowed for better matching of Sandia's phone numbers on the public network to phone numbers on the Federal Telecommunications System or FTS... a CCSA operated for the Federal Government. But we'll talk about that later. The 5ESS was also equipped with ISDN PRI tie lines to a sibling 5ESS at Sandia California in Livermore, providing inexpensive calling and ISDN features between the two sites.

This is a good time to discuss digital Centrex. Traditional telephony, even today in residential settings, uses analog telephones. Business systems, though, made a transition from analog to digital during the '80s and '90s. Digital telephone sets used with business systems provided far easier access to features of the key system, PABX, or Centrex, and with fewer wires. A digital telephone set on one or two telephone pairs could offer multiple voice lines, caller ID, central directory service, busy status indication for other phones, soft keys for pickup groups and other features, even text messaging in some later systems (like my Comdial!). Analog systems often required as many as a half dozen pairs just for a simple configuration like two lines and busy lamp fields; analog "attendant" sets with access to many lines could require a 25-pair Amphenol connector... sometimes even more than one.

Many of these digital systems used proprietary protocols between the switch and telephones. A notable example would be the TCM protocol used by the Nortel Meridian, an extremely popular PABX that can still be found in service in many businesses. Digital telephone sets made the leap to Centrex as well: first by Nortel themselves, who offered a "Meridian Digital Centrex" capability on their DMS-100 exchange switch that supported telephone sets similar to (but not the same as!) ordinary Meridian digital systems. AT&T followed several years later by offering 5ESS-based digital Centrex over ISDN: the same basic capability that could be used for computer applications as well, but with the advantage of full compatibility with AT&T's broader ISDN initiative.

The ISDN user terminals manufactured by Western Electric and, later, Lucent, are distinctive and a good indication that that digital Centrex is in use. They are also lovely examples of the digital telephones of the era, with LCD matrix displays, a bevy of programmable buttons, and pleasing Bellcore distinctive ringing. It is frustrating that the evolution of telephone technology has seemingly made ringtones far worse. We will have to forgive the oddities of the ISDN electrical standard that required an "NT1" network termination device screwed to the bottom of your desk or, more often, underfoot on the floor.

Thinking about these digital phones, let's consider the user experience of Centrex. Centrex was very flexible; there were a large number of options available based on customer preference, and the details varied between the Centrex host switches used in the United States: Western Electric's line from the 5XB to the 5ESS, Nortel's DMS-100 and DMS-10, and occasionally the Siemens EWSD. This all makes it hard to describe Centrex usage succinctly, but I will focus on some particular common features of Centrex.

Like PABXs, most Centrex systems required that a dialing prefix (conventionally nine) be used for an outside line. This was not universal, "assumed nine" could often be enabled at customer request, but it created a number of complications in the dialplan and was best avoided. Centrex systems, because they mostly belonged to larger customers, were more likely than PABXs to offer tie lines or other private routing arrangements, which were often used by dialing calls with a prefix of 8. Like conventional telephone systems, you could dial 0 for the operator, but on traditional large Centrex systems the operator would be an attendant within the Centrex customer organization.

Centrex systems enabled internal calling by extension, much like PABXs. Because of the large size of some Centrex-CU installations in particular you are probably much more likely to encounter five-digit extensions with Centrex than with a PABX. These types of extensions were usually designed by taking several exchange prefixes in a sequence, and using the last digit of the exchange code as the first digit of the extension. For that reason the extensions are often written in a format like 1-2345. A somewhat charming example of this arrangement was the 5ESS-based Centrex-CU at Los Alamos National Laboratories, which spans exchange prefixes 662-667 in the 505 NPA. Since that includes the less desirable exchange prefix 666, it was skipped. Of course, that didn't stop Telnyx from starting to use it more recently. Because of the history of Los Alamos's development, telephones in the town use these same prefixes, generally the lower ones.

With digital telephones, Centrex features are comparatively easy to access, since they can be assigned to buttons on the telephones. With analog systems there are no such convenient buttons, so Centrex features had to be awkwardly bolted on much like advanced features on non-Centrex lines. Many features are activated using vertical service codes starting with *, although in some systems (especially older systems for pulse compatibility) they might be mapped to codes that look more like extensions. Operations that involve interrupting an active call, like transfer or hold, involve flashing the hookswitch... a somewhat antiquated operation now more often achieved with a "flash" button on the telephone, when it's done at all.

Still, some analog Centrex systems used electrical tricks on the pair (similar to many PABX) to provide a message waiting light and even an extra button for common operations.

While Centrex initially appealed mainly to larger customers, improvements in host switch technology and telephone company practices made it an accessible option for small organizations as well. Verizon's "CustoPAK" was an affordable offering that provided Centrex features on up to 30 extensions. These small-scale services were also made more accessible by computerization. Configuration changes to the first crossbar Centrex service required exchange technicians climbing ladders to resolder jumpers. With the genesis of digital switches, telco employees in translation centers read customer requirements and built switch configuration plans. By the '90s, carriers offered modem services that allowed customers to reconfigure their Centrex themselves, and later web-based self-service systems emerged.

So what became of Centrex? Like most aspects of the conventional copper phone network, it is on the way out. Major telephone carriers have mostly removed Centrex service from their tariffs, meaning they are no longer required to offer it. Even in areas where it is present on the tariff it is reportedly hard to obtain. A report from the state of Washington notes that, as a result particularly of CenturyLink removing copper service from its tariffs entirely, CenturyLink has informed the state that it may discontinue Centrex service at any time, subject to six months notice. Six months may seem like a long time but it is a very short period for a state government to replace a statewide telephone system... so we can anticipate some hurried acquisitions in the next couple of years.

Centrex had always interacted with tariffs in curious ways, anyway. Centrex was the impetus behind multiple lawsuits against AT&T on grounds varying from anti-competitive behavior to violations of the finer points of tariff regulation. For the most part AT&T prevailed, but some of these did lead to changes in the way Centrex service was charged. Taxation was a particularly difficult matter. There were excise taxes imposed on telephone service in most cases, but AT&T held that "internal" calls within Centrex customers should not be subject to these taxes due to their similarity to untaxed PABX and key systems. The finer points of this debate varied from state to state, and it made it to the Supreme Court at least once.

Centrex could also have a complex relationship with the financial policies of many institutional customers. Centrex was often paired with services like WATS or tie lines to make long-distance calling more affordable, but this also encouraged employees to make their personal long-distance calls in the office. The struggle of long-distance charge accounting lead not only to lengthy employee "acceptable use" policies that often survive to this day, but also schemes of accounting and authorization codes to track long distance users. Long-distance phone charges by state employees were a perennial minor scandal in New Mexico politics, leading to some sort of audit or investigation every few years. Long-distance calling was often disabled except for extensions that required it, but you will find stories of public courtesy phones accidentally left with long-distance enabled becoming suddenly popular parts of university buildings.

Today, Centrex is generally being replaced with VoIP solutions. Some of these are fully managed, cloud-based services, analogous to Centrex-CO before them. IP phones bring a rich featureset that leave eccentric dialplans and feature codes mostly forgotten, and federal regulations around the accessibility of 911 have broadly discouraged prefix schemes for outside calls. On the flip side, these types of phone systems make it very difficult to configure dialplan schemes on endpoints, leading office workers to learn a new type of phone oddity: dialing pound after a number to skip the end-of-dialing timeout. This worked on some Centrex systems as well; some things never change.

[1] Later called US West, later called Qwest, now part of CenturyLink, which is now part of Lumen.