User: Password:
|
|
Subscribe / Log in / New account

Linux Foundation

Linux Foundation

Posted Oct 21, 2010 8:46 UTC (Thu) by alex (subscriber, #1355)
In reply to: How not to recognize free hardware by eru
Parent article: How not to recognize free hardware

I would like the LF to come up with a mark like:

Works with 2.6.32+

Which would mean all the hardware had mainline (not staging) support from that kernel onwards.


(Log in to post comments)

Linux Foundation

Posted Oct 21, 2010 10:06 UTC (Thu) by cortana (subscriber, #24596) [Link]

That is a great idea and I'd love to see it implemented. However, care would need to be taken with hardware that depends on the presence of firmware that is bundled with the kernel, without a license to do so, and then stripped out by distributions such as Debian and Fedora.

The firmware problem...

Posted Oct 21, 2010 11:33 UTC (Thu) by alex (subscriber, #1355) [Link]

Do the kernel people still accept firmware blobs into the tree?

The firmware problem...

Posted Oct 21, 2010 13:30 UTC (Thu) by BenHutchings (subscriber, #37955) [Link]

Not outside of staging, no.

From that kernel onwards ?

Posted Oct 21, 2010 10:18 UTC (Thu) by copsewood (subscriber, #199) [Link]

I think it improbable that a software development organisation would be able to guarantee that all future versions of a program would be able to work with a specific item of hardware indefinitely. Eventually you have to have the freedom to break backwards compatibility with outdated stuff.

For a trivial example, consider the case of a single chip computer including non expandable RAM. 2.6.X might work in this amount of RAM, 2.6.X+1 might not.

From that kernel onwards ?

Posted Oct 21, 2010 11:38 UTC (Thu) by alex (subscriber, #1355) [Link]

I think the definition would have to be a little grown up in accepting realistic caveats.

Certainly for most hardware that is working from a given kernel version you can reasonably expect it to continue working until the hardware is becoming a museum piece. In the cases that don't you know it worked for at least one version of the kernel and if you care enough (or pay someone to care) you could get it working again. This is a mark about freedom of hardware not assurance of eternal support.

Maybe just drop the + from the mark?

It is a contest?

Posted Oct 22, 2010 15:38 UTC (Fri) by khim (subscriber, #9252) [Link]

FSF's idea was already DOA, why do you want even crazier and more pointless one?

At least with "Respects Your Freedom" mark you have a clear list of requirements. "Works with 2.6.32+" does not have such a list - because there are no clear-cut rules which can be followed to guarantee inclusion in linux kernel!

This is idea will be workable if and when all devices will use standardized interfaces (a-la USB sticks), but
1. It does not look like it'll happen any time soon and
2. If this will actually happen the mark will be even more pointless: all devices with happily work with Linux at this point.

It is a contest?

Posted Oct 22, 2010 17:14 UTC (Fri) by tialaramex (subscriber, #21167) [Link]

Doesn't it feel like this is happening more & more?

I admit I don't have hard figures, and I'm not sure how I'd get them, but my strong feeling is that things have been getting progressively better.

That is to say, more of the hardware the OS actually cares about is class compliant rather than requiring custom drivers.

There are big holdouts. Networking for example, Ethernet cards are hot on compliance with a well documented standard right up until you reach the controller chip, I guess all the standard-friendly people at network peripheral companies work over on the link-layer side of the operation.

But on the whole I can't think of any category where things are less class-based than they were in say 1995, and many where things are more so.

I thought we'd missed a trick twice with USB, firstly sound (but today everything is either USB Audio or HDA on PCI) and then webcams (but the USB video spec didn't die, it was just a long pregnancy and today you can walk into any store and buy webcams you've never heard of that work in Linux)

You are correct - that's how things are going...

Posted Oct 24, 2010 20:06 UTC (Sun) by khim (subscriber, #9252) [Link]

That is to say, more of the hardware the OS actually cares about is class compliant rather than requiring custom drivers.

Well, the situation is simple: most "hardware" devices nowadays include sizable software components (firmwares for things things like HDD today are bigger then whole Unix systems were 30 years back), so when it's known what the device should do and how it should work they eventually comply to the spec.

But it does not make the suggested mark valuable! In first stage when new wave of the devices are introduced it's pointless because it's sheer insanity to tie release dates for the "hot new stuff" to release dates of kernel and later when majority of the devices are produced by the most famous brand ever called "Noname" there are noone who will care enough to bother to test compliance - but it does not really matter because stuff works out of the box!

But on the whole I can't think of any category where things are less class-based than they were in say 1995, and many where things are more so.

Any category which existed back then is "old news" today - so of course the majority of the devices reached "driver for class" stage. But new categories and subcategories are different. Compare situation with mices (some still require drivers to enable full functionality, but for majority there are not need for driver at all) and multitouch-enabled touchpads.


Copyright © 2017, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds