"Tom Van Baak" <[EMAIL PROTECTED]> wrote: > If a device is USB-only you pretty much have to just plug it > into a wintel PC, install their OS-dependent software, and > take what you're given.
There's not much in the way of OS-dependent software, at least not in the conventional sense of "must install custom driver or it won't work." USB serial adapters of the FDTI ilk just work in Windows and Linux. (Yes, a manufacturer can assign its own ID, and that will require a custom wrapper for the standard driver ... but it doesn't happen that much, as it is ordinarily a bad idea.) Serial usb devices -- most USB devices in fact -- work quite well on non-x86 hardware, too. I can tell you from experience, mostly with ARM platforms (XScale and StrongARM). (I worked on the OHCI usb host controller driver for the StrongARM SA1110 for Linux. Don't get me started on the silicon bugs in this part.) "Poul-Henning Kamp" <[EMAIL PROTECTED]> wrote: > The biggest mistake in USB, was that they didn't define mandatory > device models, like for instance SCSI. What are you talking about? The standard USB Mass Storage model can directly encapsulate and transport SCSI commands. In fact that's about all it can do. The entire spec is only 7 pages, because it is layered on top of all the T10 specs. Furthermore, there are a number of other standard device models: hid, cdc, audio, et al. There's an entire working group (DWG) for this. Arguably there are few ad hoc device interfaces that ought to be standardized, but the state isn't dire. I look at that as the market at work. (Compare this situation to that of Bluetooth. There are several half-baked and useless device profiles in the specs that were defined too soon in the specification process, before experience with the devices could be considered and reflected back into the specs.) -ch _______________________________________________ time-nuts mailing list time-nuts@febo.com https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts