LWN.net Logo

ABS: Android beyond the phone

ABS: Android beyond the phone

Posted Apr 21, 2011 16:50 UTC (Thu) by iabervon (subscriber, #722)
Parent article: ABS: Android beyond the phone

I'm not sure it makes sense to have kernel drivers for the embedded hardware at all, let alone have them in the mainline kernel; these aren't drivers for a printer, they're drivers for a collections of motors, valves, and sensors arranged to be useful for printing. I think it would make most sense to have a userspace daemon that knows how to use the particular arrangement in this box to print things, and have the kernel consider it all a bunch of uninteresting GPIO. (Uninteresting in that it doesn't even reflect or affect kernel state, so it's less relevant to the kernel than a caps lock LED or a battery voltage indicator.) This would also be the key to allowing the kernel (and other userspace programs) to be replaced separately from the device-specific stuff. Sure, someone might be able to improve the quality of the printing by replacing this code, but it's orthogonal to improving the network stack or UI. It's important that I am able to change some code to be able to solve my problem; it's even more important that solving my problem doesn't require me to change all of the code that there is.


(Log in to post comments)

ABS: Android beyond the phone

Posted Apr 21, 2011 17:14 UTC (Thu) by w_sang (subscriber, #52415) [Link]

It is not about "if it makes sense" (besides, I think it does). To the best of my knowledge, vendors will in most cases simply have to ship their changes to the kernel. The question to the speaker was if those vendors are aware of this and are prepared to have the necessary infrastructure. The fact, that the even he, being a consultant as far as I understood, did not (or did not want to?) understand the question gave a good picture of the current situation IMO.

Why should they ship anything

Posted Apr 22, 2011 0:21 UTC (Fri) by khim (subscriber, #9252) [Link]

Well, the literal reading of the GPL will of course mean that drivers should be shipped but when there are literally hundreds of millions of devices where drivers and binary-only blobs (I mean routers)... will any court really accept the theory that it was not the intention of developers in first place - what with EXPORT_SYMBOL_GPL and everything?

Why should they ship anything

Posted Apr 22, 2011 3:41 UTC (Fri) by mjg59 (subscriber, #23239) [Link]

The point of EXPORT_SYMBOL_GPL has always been to indicate that the authors feel that there's no possible way that you could use the functionality without being a derived work. But it's also always been explicit that this is not intended to mean that avoiding EXPORT_SYMBOL_GPL means that you're *not* a derived work. At that point you're just in a slightly greyer area.

Copyright © 2013, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds