Linked by Thom Holwerda on Tue 21st Feb 2012 13:50 UTC, submitted by Ducky Johnson
ReactOS Now this is what I like to see. Smaller, alternative operating systems have been dying by the dozens these past 6 years or so, so in order for the remainder to survive, they need to work together. ReactOS and Haiku have been doing that for a while now, and the latest fruits of this collaboration is a much-improved USB stack for ReactOS.
Thread beginning with comment 508040
To read all comments associated with this story, please click here.
No small feat
by weland on Tue 21st Feb 2012 22:05 UTC
weland
Member since:
2012-02-19

As someone else mentioned above, this is no small feat.

Writing a USB stack is *really* hard. I helped a colleague who was working on a client (not host!) USB stack for the MSP430. Even with sample code and some support from TI, it took him almost a month to implement a subset of the ACM functions that we needed. I don't even want to think about what writing the USB host stack involves.

Reply Score: 3

RE: No small feat
by pgeorgi on Wed 22nd Feb 2012 07:51 in reply to "No small feat"
pgeorgi Member since:
2010-02-18

I think host support is easier.
Since in USB all communication is initiated by the host, the driver author gets to decide what to do and when.

For clients, you'll have to cope with whatever the host throws at you. (but to be fair: some devices are rather weird in what they answer)

But I only wrote a bunch of host drivers (ohci, uhci, ehci, some xhci stub, various class drivers) so far, and no client drivers.

Reply Parent Score: 1

RE[2]: No small feat
by weland on Wed 22nd Feb 2012 11:23 in reply to "RE: No small feat"
weland Member since:
2012-02-19

In all fairness, coping with weird input was indeed difficult, but we still only had to support a single class and we could cut some portability corners due to the low resources (8K of RAM and 48K of flash is enough but isn't *that* much).

Reply Parent Score: 1