The ins and outs of USB (IBM developerWorks) [LWN.net]

Peter Seebach discusses the USB standard on IBM developerWorks. The article is mostly presented from a Window/Mac perspective. "The USB specification may be an example of that hybrid de jure or de facto standard, one that clearly earned wide acceptance through its technical merit. Learn the history of the USB standard and some of its benefits to users and vendors, as well as where it missed the boat."

to post comments

The ins and outs of USB (IBM developerWorks)

Posted Apr 28, 2005 22:20 UTC (Thu) by njhurst (guest, #6022) [Link] (26 responses)

If USB won on technical merit then what happened to firewire? Firewire seems to be superior in every way (speed, power, flexibility, support for interrupts etc). Perhaps USB actually won because of market reasons? This article seems somewhat revisionist.

Cheaper == Better :)

Posted Apr 28, 2005 22:56 UTC (Thu) by AnswerGuy (guest, #1256) [Link]

Low price offers huge "technical merit."

JimD

The ins and outs of USB (IBM developerWorks)

Posted Apr 28, 2005 23:01 UTC (Thu) by drathos (guest, #6454) [Link] (15 responses)

I'm guessing (read: hoping) that he's referring to it 'winning' over the technologies it replaced (and even then, I'm not sure if it were merit or brute force that got it where it is). PS2, serial, and parallel connections have all been replaced by USB to a certain extent.

Yes, you still see those connections on motherboards, but it's getting harder and harder to find uses for them. You'll be hard pressed to find parallel printers for sale anymore. Most mice are USB with PS2 adapters now. Just about everything that used serial before uses USB now. Keyboards are the only thing that seems to be sticking with the old connectors.

The ins and outs of USB (IBM developerWorks)

Posted Apr 29, 2005 6:39 UTC (Fri) by beejaybee (guest, #1581) [Link] (14 responses)

Yeah, designed-in obsolescence. There is _nothing_ wrong with PS/2 connectors for keyboard & mouse, indeed there seems to be no benefit in USB for these devices. RS-232 serial mice and AT keyboards are fine, too. I don't see any performance benefit in USB over parallel ports for printers, either. Meanwhile parallel port connectors are solid and reliable, USB connectors aren't.

It will really upset me when PS/2 connectors get dropped from mobos. You see, I'm using a 20-year-old IBM keyboard with a nice positive action, decently weighted and with a nice "click". It has a PS/2 connector. I find it next to impossible to use membrane keyboards due to their "dead flesh" feel; the only halfway acceptable modern replacement is a mechanical contact keyboard from Cherry (at 20x the price of a "standard" kb), but it doesn't have the same weighted feel or the same audible feedback.

As for USB storage - well, it does seem to work to some degree, but it's infuriating in some ways (devices that work in some ports but not others). I definitely agree with others who claim that Firewire is superior from the practical point of view, never mind the technology.

This is essentially the same as the digital/film photography argument. The market is removing choice on the principle that sales volume rules, therefore we all get "cheap and cheerful" devices that are infuriating to use, unmaintainable and will need replacement in months rather than decades.

The ins and outs of USB (IBM developerWorks)

Posted Apr 29, 2005 6:56 UTC (Fri) by donwaugaman (subscriber, #4214) [Link]

They have PS/2 mouse + keyboard -> USB adapters for about $10. Obsolescence delayed. :-)

The ins and outs of USB (IBM developerWorks)

Posted Apr 29, 2005 6:58 UTC (Fri) by alspnost (guest, #2763) [Link]

Yup, don't get me started on keyboards. I have a USB keyboard, which is great, but when you buy a new motherboard, USB keyboard support always seems to be *disabled in the BIOS* by default. So what do you do? Go and steal your neighbour's old PS2 keyboard, use it to get in to the BIOS to change one parameter, then plug in your USB keyboard again. Ridiculous - why do they do this??

IBM keyboard

Posted Apr 29, 2005 8:35 UTC (Fri) by grmd (guest, #4391) [Link]

the only halfway acceptable modern replacement is a mechanical contact keyboard from Cherry

Do you know about Unicomp? They sell what is essentially the IBM clicky keyboard from their pckeyboard website. I have bought both the clicky and spongy types and been happy with them. The clicky type feels nicer but the noise can be annoying to co-workers, so sometimes I use the spongy one instead, and even that feels better than the cheap modern keyboards that most people seem to use.

The ins and outs of USB (IBM developerWorks)

Posted Apr 29, 2005 10:50 UTC (Fri) by bert.kenward (subscriber, #28573) [Link] (5 responses)

Hot plug.

The ins and outs of USB (IBM developerWorks)

Posted Apr 29, 2005 13:35 UTC (Fri) by smitty_one_each (subscriber, #28989) [Link]

Plus, it's cheaper to group all of the I/O USB stuff on one USB hub.

The ins and outs of USB (IBM developerWorks)

Posted Apr 30, 2005 18:39 UTC (Sat) by bk (guest, #25617) [Link] (3 responses)

PS/2 is just as hotplug as USB for keyboards and mice, at least on Linux. Try it. Unplug your PS/2 keyboard while the system is running and then plug it back in. Everything works.

The ins and outs of USB (IBM developerWorks)

Posted Apr 30, 2005 20:54 UTC (Sat) by farnz (subscriber, #17727) [Link] (2 responses)

PS/2 hotplug blows a fuse on my motherboard. YMMV, but it's risky.

The ins and outs of USB (IBM developerWorks)

Posted May 1, 2005 2:43 UTC (Sun) by dlang (guest, #313) [Link] (1 responses)

it varies motheboard to motherboard, but I can definantly attest that hotplugging PS/2 keyboards can pop fuses on the motherboard (it depends which pins connect first, just the right combination will cause problems)

back when I worked PC repair I repaired dozens of machines with this exact problem (interestingly the manufacturers expected service was motherboard replacement)

The ins and outs of USB (IBM developerWorks)

Posted May 1, 2005 15:09 UTC (Sun) by bk (guest, #25617) [Link]

Weird. It's worked fine on all my machines, I didn't even know it could cause problems. Nice to know, I guess.

The ins and outs of USB (IBM developerWorks)

Posted Apr 29, 2005 14:17 UTC (Fri) by hensema (guest, #980) [Link]

USB Mice use higher bitrates than PS/2 mice. You can actually notice the difference. Also, the PS/2 mouse port uses an interrupt, and interrupt lines are still a bit scarce.

The ins and outs of USB (IBM developerWorks)

Posted Apr 29, 2005 16:14 UTC (Fri) by josh_stern (guest, #4868) [Link]

I feel the same way about keyboard style. As someone else mentioned, you can buy an adapter. Also, I've found the Macally USB keyboard to have a relatively good spacing and feel.

The ins and outs of USB (IBM developerWorks)

Posted Apr 29, 2005 17:30 UTC (Fri) by iabervon (subscriber, #722) [Link]

There is _nothing_ wrong with PS/2 connectors for keyboard & mouse, indeed there seems to be no benefit in USB for these devices.

Unless you have configurable input devices, such as the Twiddler or Kinesis. With USB, the computer can configure them (load your saved settings, etc); with PS/2, you have to type your configuration into the device.

The ins and outs of USB (IBM developerWorks)

Posted Apr 29, 2005 23:49 UTC (Fri) by wildpossum (guest, #17744) [Link]

>I don't see any performance benefit in USB over parallel ports for printers, >either. Meanwhile parallel port connectors are solid and reliable, USB >connectors aren't.

You haven't tried to send a multi-megabyte PostScript file to a printer over a parallel port then. I'd prefer an Ethernet internet on my printer but I'll settle for USB 2.0.

The ins and outs of USB (IBM developerWorks)

Posted May 2, 2005 6:14 UTC (Mon) by ekj (guest, #1524) [Link]

Yeah, designed-in obsolescence. There is _nothing_ wrong with PS/2 connectors for keyboard & mouse, indeed there seems to be no benefit in USB for these devices.

I consider it a "wrong" that the fuse on my motherboard blows if I hotplug my keyboard.

I consider it a "wrong" that the communication is one-way, so that I cannot program my mouse over ps/2.

I consider it a "wrong" (or atleast a minus) to require multiple different cables instead of just a single one to the computer under the desk and a hub at the desk.

For that matter, I consider it a "wrong" to have different connections for situations where there's no reason you couldn't use one and the same. It decreases flexibility for no reason. It assumes too much. (Everyone will have precisely *one* parallell device for example. Yes, there's pass-troughs but it's a hack.

Parallell is ridicolously thick and large and at the same time slow as molasses. On Half the motherboards, the small screws like to fall off if you look at them funny.

The ins and outs of USB (IBM developerWorks)

Posted Apr 29, 2005 11:56 UTC (Fri) by Incabulos (guest, #29637) [Link] (5 responses)

I take it you have not come across firewire devices? I was using an external 200GB harddisk and dvd/rw drive under Linux today, both of which are firewire drives. Even on a distribution as old as RedHat9 it was a matter of plugging it in, fdisking the new /dev/sd* device that appeared, running mkfs, and it was good to go, looking for all the world like a local SCSI disk. If only everything was so simple.

The ins and outs of USB (IBM developerWorks)

Posted Apr 29, 2005 12:50 UTC (Fri) by njhurst (guest, #6022) [Link] (4 responses)

I'm sorry, I don't understand what this has to do with my question? I was merely pointing out that firewire, which is superior in every(?) respect is almost a dead standard, only holding out in the top end (much like scsi). Why can't I buy a 4GB/s optical firewire drive to go in my desktop case, plugging straight into the firewire port on the mobo? Instead we have USB, USB 2, SATA, PCI-E etc. Technical merit is clearly not the main factor in USB's acceptance (I suspect that Intel being piggy is more to do with it?).

I certainly have used firewire devices, Indeed I was one of the first users of firewire storage (having an early firewire apple machine running linux/ppc).

The ins and outs of USB (IBM developerWorks)

Posted Apr 29, 2005 15:44 UTC (Fri) by kakareka (guest, #5204) [Link] (3 responses)

IIRC, around the time firewire had become an IEEE standard and was gaining acceptance, Apple decided to announce it was going to charge royalty fees for its use. Intel and other manufacturers were rather upset over that and decided to throw their support to USB2 instead. At that time, I believe pratical implementations of firewire offered significantly less bandwidth than what it theoretically could achieve, and not much higher than what USB2 was going to deliver. Apple rescinded the royalty fees (or dropped them down pretty significantly) a few months later, but the damage was done.

As for the various bus types, they generally target different markets/apps. For instance, AFAIK, PCI-E is really intended for board-board connections rather than cable connections. It really isn't a competitor to firewire although things like hard drives bridge the gap. SATA is designed for internal devices, so has much shorter cable lengths which means a lot for the electrical design (rise/fall times, propagation delays,etc).

John

The ins and outs of USB (IBM developerWorks)

Posted Apr 29, 2005 18:11 UTC (Fri) by jwb (guest, #15467) [Link] (2 responses)

The premise of your post is incorrect. Apple never did and never could charge royalties on the use or implementation of the IEEE 1394 bus. What it proposed was to charge royalties on their trademarks, the FireWire name and logo. I believe the proposed charge was $1 per device.

The ins and outs of USB (IBM developerWorks)

Posted Apr 29, 2005 18:42 UTC (Fri) by kakareka (guest, #5204) [Link] (1 responses)

Doing a quick google search does seem to confirm it was a patent issue, not a trademark one. Apple does own patents on the technology, and you can have patents (and charge royalties) on technology included in IEEE standards. Appears other companies claimed patents as well, and forced Apple to agree to cross license, which is what I believe usually happens with IEEE (and other organizations) standards anyway.

The ins and outs of USB (IBM developerWorks)

Posted Apr 29, 2005 18:51 UTC (Fri) by maxmtl2002 (guest, #3990) [Link]

For the 1394 licensing thing:

http://www.mpegla.com/1394/

Referred from one of the main 1394 site:

http://www.1394ta.com/

Looks like the patents cost are US$ 0.25 for each device. I do not know about its historic values.

The ins and outs of USB (IBM developerWorks)

Posted Apr 29, 2005 15:53 UTC (Fri) by cott (guest, #6931) [Link] (2 responses)

I'm not convinced USB has "won". They have different purposes.

USB is for devices you used to hook up with serial or parallel ports: modems, printers, keyboards, and mice. Do you really think firewire is appropriate for those things?

Firewire is for peripherals that used to be hooked up to SCSI and/or need high-speed connections: disk drives, cameras, etc.

The ins and outs of USB (IBM developerWorks)

Posted Apr 29, 2005 21:05 UTC (Fri) by rfunk (subscriber, #4054) [Link] (1 responses)

You haven't heard about USB 2.0?

The ins and outs of USB (IBM developerWorks)

Posted Apr 29, 2005 22:30 UTC (Fri) by Robin.Hill (subscriber, #4385) [Link]

You haven't used USB 2.0? The processor load for high speed USB 2.0 is _way_ higher than for firewire (400 or 800) - more than a couple of devices running simultaneously and you're stuffed.

The ins and outs of USB (IBM developerWorks)

Posted Apr 28, 2005 23:21 UTC (Thu) by jwb (guest, #15467) [Link] (2 responses)

The article rambles. The author should have addressed the real weaknesses of USB, namely, very bad implementations, and power. USB is marred by a proliferation of fourth-rate implementations. Many USB hubs simple do not work correctly. For example, if I plug my smartcard into the powered USB ports on my Dell monitors, the card reader refuses to work. It works fine when plugged into other hubs or into the root hub. You just can't know in advance if the latest random taiwanese USB gadget will work.

Power is the other big problem. USB simply doesn't provide enough of it. IEEE1394 specifies 1.5A at 12V per port, while USB provides 500mA at 5V on certain weekdays, with a stiff tailwind, under a full moon. If you plug an iPod into a computer with a USB cable, the battery will be quickly discharged. If you use FireWire instead, the battery will be swiftly *charged*. End-users appreciate the difference. USB often fails mysteriously, either because the user plugged a current-hungry device into an unpowered hub, or because the aforementioned random taiwanese USB port is simply incapable of supplying the needed current. There exists considerable confusion in the public about powered, unpowered, bus-powered, and self-powered USB hubs.

Another considerable nuisance of USB is its total inability to provide isochronous end-to-end transmission. The interfaces -- hardware and software -- simply do not expose the needed functionality. I have recently acquired a multichannel digital audio interface supporting 24-bit, 192kHz data rates. It uses FireWire because that protocol has rich isochronous capabilities. The interface provides satisfactory low-latency, high-bandwidth performance. It couldn't have been done with USB; USB just doesn't allow for high-performance isochronous transfer.

Clearly USB is an ideal interface for cheap devices like keyboards, mice, joysticks, blinking LEDs, and so forth. But that doesn't explain the continuing existence of USB2.

The ins and outs of USB (IBM developerWorks)

Posted Apr 29, 2005 4:01 UTC (Fri) by beoba (guest, #16942) [Link]

It sounds like you have more issue with USB itself than with the article. Not that I disagree, few things are more irritating than attempting to plug in a USB keyboard when you can't see the socket.

Printing and scanning is a bigger problem

Posted May 6, 2005 12:24 UTC (Fri) by rwmj (subscriber, #5474) [Link]

I find a bigger issue is lack of standardisation for the protocols for printing and scanning.

I can plug in just about any storage device (camera, memory card, phone, ...) and it appears as "USB mass storage".

Why can't I do the same for printers and scanners? It's not as if printers and scanners differ from each other so much (rectangular page, the only major difference is the resolution). Yet I need to install special drivers for every type of printer. Certain printers (esp. HP crap) the drivers are very invasive, whatever platform you're on.

Rich.