Sunday, June 27, 2021

The parallel universe of FireWire hubs

I like FireWire and I still use FireWire (I've even used it to power a WiFi-to-Ethernet connector on a PowerBook G4), but this is a retrocomputing blog, and for the larger consumer market IEEE 1394 is now just an odd little niche. Many postmortems have been done on the death of FireWire and it still pops up in weird places like the military (see AS5643 and descendants like AIR5654A), low-latency audio, infotainment systems and some security and monitoring devices, but I think the biggest thing that doomed it was that it was perceived as a competitor to USB and failed to sufficiently differentiate itself. Device manufacturers didn't help: with the exception of some high-end A/V equipment and tape camcorders, the same devices (mass storage, webcams, scanners) showed up with USB ports as did with FireWire ports, and there were many commodity PCs that lacked FireWire entirely and many devices that were USB-only, so USB connectivity won out. Licensing costs no doubt played a major role but market perception greatly hastened the process. FireWire still has infrastructure advantages in topology, latency and segment length but it also makes devices more expensive, and even these points in its favour are outweighed by the comparatively prodigious peak bandwidth in USB 3 despite FireWire winning the bandwidth war handily for many years.

One example of this was the parallel universe of FireWire hubs. If you think of FireWire as "a big USB" then a hub wouldn't seem so strange, but FireWire was actually meant to replace SCSI. SCSI and FireWire are peer-to-peer: any device on the bus can talk to any other device, unlike USB where each bus has at most one host and the host does all the initiation of data transfer. (USB On-The-Go still has one host and one host only; it just allows certain devices like your mobile phone to swing both ways.) The point-to-point capabilities of USB 3 notwithstanding, a USB hub has one upstream port for the host and multiple downstream ports for the devices. A FireWire hub, however, is like getting a longer internal SCSI cable; more devices simply exist on the same bus. Connecting multiple FireWire hubs just makes a bigger bus because all the ports are the same.

However, this difference was mostly lost on consumer users because a FireWire hub, possibly intentionally, superficially resembled an oddball USB hub with weird connectors and a higher price. And both existed largely for the same reasons, first and foremost to give you more ports, and secondarily also to inject power into the system where multiple bus-powered connections could make voltage sag. (They could also improve stability by smoothing out impedance mismatches between controller chipsets, which solved an issue with my Sawtooth G4 freaking out when I connected two FireWire drives to it directly, even though they were both self-powered and the effective topology with a hub was still the same.) Nevertheless, FireWire hubs were absolutely distinct devices and at least in some respects catered to different markets. So what did this parallel universe look like?

Belkin was probably the most prolific manufacturer of consumer-level FireWire 400 hubs. These were targeted mostly at Mac users, though they would work just fine with PCs. Their first product was the F5U524, which was a four-port powered FW400 system (one port connected to the computer, so you got three extra ports). They sold far more of the F5U526 6-port hubs, though, and the most common form factor was this one:

Yes, that particular unit was a Fry's reject that eventually ended up in the Weird Stuff bargain bin (rest in peace), but the price tag gives you an idea of why consumers might have found it unappealing. It also came in silver, shown here. I have a silver F5U526 connected to my Quad G5 and my Sawtooth G4.
Belkin later made a awkward curved variant with one port in the front and five in the back. Although still marked as a F5U526, it was only rated for 1.25A instead of 1.5A.

Nearly as prolific was Kramer Tools, who still specialize in high-end A/V hardware. In addition to their rack-mount FireWire switchers (controllable by IR remotes or RS-232/485) and FireWire active repeaters, they made four-, six- and even eight-port powered FW400 hubs intended for prosumer and professional studios. These were definitely not Best Buy specials. Made of durable steel, you'd probably fracture someone's skull if you lobbed it at them. Their eight-port FW400 was the biggest FW400 hub I could find advertised.

A similar consumer device to the Belkins was IOGear's GFH610 "FireWire Hud" [sic]. This was a powered 6-port FW400 unit, five in the rear, one in the front. The label pic was stolen from an eBay auction (not mine, not affiliated); hope they don't mind.
The label typo inspires confidence. If you had only half the confidence, the GFH310 only gave you half the ports (two in the rear, one in the front).

Potentially the most outrageous FireWire hub (thanks jonahjoselow) was the Hubzilla. It's like it sounds: Godzilla with four FW400 ports embedded in his spine. I would destroy cities too if I were subjected to this sort of torture. (Image from PCWorld.)

It had an MSRP of $75, an external power option and LEDs for connectivity state and wrecking Tokyo. However, it's not clear if any actually got sold.

As a degenerate case some FireWire devices offered daisy chaining with input and output ports, which could be considered a two-port hub of sorts (just one port being the internal device); the Lexar Professional Compact Flash card reader (shown here) and the Iomega Minimax are examples.

I remain a big fan of the Lexar FW CF reader, which transferred data from my camera to my Quad G5 in record speed and certainly quicker than any USB 2.0 card reader. Compared to even the USB 3.0 reader on my Talos II, it's still no slouch. However, the Lexar relies on bus power, so it's a suboptimal hub for that purpose.

As FireWire devices became less common, some manufacturers perceived a market for dual hubs — i.e., USB 2.0 and FireWire hubs in the same physical unit. These were not complicated devices technologically because they weren't anything more than two devices on one board in one case, with a plug each for the USB and FireWire sides, and they don't convert USB to FireWire or vice versa. All of the combo hubs I've personally seen have four USB ports and three FW400 ports, like the IOGear GUH420:

Note, however, that you really do get four extra USB ports; devices of this sort had an additional B-jack to connect to the host computer and presented four USB A-ports for downstream devices. On the FireWire side, though, and like the FW-only units, one of the FireWire ports had to connect to your computer and thus you only got two extra FW ports (not three). This was not a good look for comparison purposes.

Additionally, unlike the FW-only hubs which universally offered an external power option, some of these were powered and some of them weren't, and almost all of the ones that were powered only seem to power the USB side (the tipoff is the wallwarts only provide 5 volts, not 12). The same basic notion was used by devices from Hewlett-Packard (targetted at Windows PCs, with a micro-USB jack for the PC connection) and zombie-Compaq after HP bought them. SIIG sold an internal USB-FW hub for 3.5" drive bays with a 4-pin FW400 and a 6-pin FW400 (plus the internal 6-pin FW400 to connect to the motherboard), but those are electrically identical modulo the power pins, so they're really the same sort of device. It had a connection for 4-pin motherboard power, which because that has a 12V rail, was one of the few devices to power both sides of the hub.

Belkin also made the F5U507, an external dual device for the Mac market which was also unpowered:
An interesting variation on the 4-3 port design was the Moshi iLynx. With all the ergonomics of an overgrown doorstop, it offered one FW400 and two USB 2.0 ports on each side of the "hump," and thus only advertised itself as a 4-2 device. However, it's actually the same as the others here because it has male FireWire and USB plugs (not jacks) for the host exiting the back. This means it's still electrically a "4-3," but at least it was honest about how many more ports you really got. It was also not powered.
FireWire 800 was a big jump for mass storage performance and kept the protocol relevant for awhile longer, even though for video applications old brown FW400 was still just dandy. By then, however, the market was much smaller and most of the hub options either evaporated entirely or moved prosumer. Belkin did not make an FW800 hub, dual or otherwise, though their Thunderbolt (and other Thunderbolt) hubs have a single FW800 port. Moshi upgraded the iLynx to the iLynx 800, but was otherwise the same form factor:
Kramer Tools also expanded their offerings to FW800, though they only produced three and four port varieties (there was also a two-port which is properly considered a repeater as it's designed to connect to another repeater over RG-6 coax). If you wanted an eight-port FW800 hub, then you had to have the Nitro AV. I don't have eight FW800 devices, but I still had to have it:
This monster comes with a 3A power supply, more than enough to keep every power-hungry beast on the bus docile and purring:
Some companies are still producing, or at least still make available, FireWire 800 hubs. Amazon lists a no-name 3-port device:
I don't know anything about the manufacturer and it seems a lot of people think it's junk.

And yes, I'm still using FireWire, but only for the same specific purposes I used it for back in the day. Besides A/V (my HDV roadgeek camcorder, my Canopus ADVC-300 framegrabber and my original iSight), I almost entirely use FireWire for booting and exchanging files with the Power Macs. But it lingers in prosumer A/V applications like recording studios and multicamera security setups, and big multiple-endpoint configurations like that are where FireWire's topology really shows off its flexibility. If you're in that category and also absolutely nucking futs, I spotted this auction for a terrifying 16-port FW800 device in an external PCIe enclosure (not my auction, not affiliated).

Put enough juice through it and it might even start a FireWire fire. Try that with USB. Just don't blame me if you succeed.

Monday, June 7, 2021

Monterey? BTDT. Try Project Monterey.

Apple's announcement of the next version of macOS, Monterey, means my 2014 MacBook Air now gets to join my Quad G5 in the "not supported" category (not that I care, it's Mojave Forever). But it's a good reminder of the previous Project Monterey, a multicorporation attempt to make the One Unix To Bind Them All from IBM AIX, SCO (then the putative holders of the True Unix) and DYNIX/ptx that would run on the One Architecture To Bind Them All, IA-64 (a/k/a Itanium). In case you weren't yet with the new hotness, it would run on your old and busted 32-bit x86 hardware, too.

Today you'd laugh your fool head off at the very thought of "Itanic" taking over the world, but when it was announced in October 1998 Monterey was a credible threat. By having IBM, SCO, Sequent (which IBM bought) and Intel as backers its ascent to dominance seemed inevitable, and its ability to run on existing and future hardware along with the jackboots of AIX and the multiprocessing strength of Dynix was thought to be strongly appealing to high-end enterprise IT. (The issue of IBM's then-contemporary POWER server line and Intel's Pentium server offerings potentially in direct competition was handwaved away.) A long list of the usual hangers-on backed it at the time as well, including Acer, Compaq, Groupe Bull, Samsung and Unisys.

The damn thing actually shipped, too, because most of it was based on already extant code. Project Monterey's first release was to essentially repackage AIX on POWER, and UnixWare 7 and DYNIX/ptx on x86; in 2000 the next wave and the "real" Project Monterey was AIX 5L for IA-64, which IBM actually sold on request and apparently had some small number of running systems in the wild.

Oddly, what doomed Project Monterey was Linux on IA-64, the so-called "Trillian Project" that emerged in mid-1999. Intel, always one to hedge its bets, was part of that effort along with Silicon Graphics, VA Linux and Hewlett-Packard, but most of the work was done by Cygnus before their eventual purchase by Red Hat. SGI and HP, of course, made their own Itanium machines; HP, to its current chagrin, still does. As if in response IBM promised Monterey would have strong Linux compatibility, but if you needed Linux compatibility as a primary feature, why not just run Linux? A Caldera executive was quoted in InfoWorld that year saying, "I would expect over the next one to two years [Linux for IA-64] will catch up and in some cases exceed Monterey, for no other reason than the sheer number of people contributing to Linux."

And, well, that's exactly what happened; Itanium outlasted Monterey, and Monterey went down in flames. IBM sold less than 50 licenses by the time Monterey was quietly shot in the head in 2003, though some sources say IBM had already pulled out as early as 2001. Its breakdown directly led to the SCO vs IBM lawsuit in which SCO went bankrupt and in a related case was found never to have had the Unix license to grant to IBM in the first place. Itanium, for its part, will cease shipments a little over a month from now on July 29, 2021.

Somehow I just don't see this Monterey being that interesting.