Comment 13 for bug 812394

Ted Gould [2011-08-18 20:27 -0000]:
> I guess the point here Martin is that you're saying "we don't know" --
> and that leads to very confused designs that say we don't know to the
> user, and thus the user doesn't know. Then they have no confidence in
> us to tell them anything!

I'm afraid I don't understand what you mean here by "us tell them
anything" -- you mean if we had a whitelist of certified models, and
suspend would still fail because of a special case/USB device/etc.,
they would not know?

> I think that we need to come up with a "yes" or "no" answer to the
> question: "Does suspend work?"

Depends what "yes" is -- "always, 100%" is probably "no" everywhere,
as both hardware and to a bigger degree software have a tendency to
deterioate after some time of running (bugs corrupting the memory
state, memleaks), and we can't ever test suspend with all USB or
bluetooth hardware plugged in in the world.

I think "yes" should be "good and reliable enough to be useful", which
seems to be the case on most platforms these days? This could be
measured with some synthetic test cases like "survives 20 suspends in
a row after a clean boot" and "survives a 24 hour suspend cycle".

> Or "yes, with a blacklist" or "No, with a whitelist" would also work.

A blacklist might help, but I don't think that the immense effort in
keeping it up to date justifies the little gain that you get from it
(it would be better to spend that time figuring out pm-utils quirks to
fix it, etc.)

> Do you think this would be a good thing to put on the technical board's
> agenda?

If John and/or you still want to disable suspend by default, then the
TB is the appropriate forum to decide about this, indeed.

Thanks,

Martin

--
Martin Pitt | http://www.piware.de
Ubuntu Developer (www.ubuntu.com) | Debian Developer (www.debian.org)