[LAU] The Many Ways of Pam Limits...

[Jack O'Quin]

On Mon, Jul 20, 2009 at 10:33 AM, Paul Davis<paul at linuxaudiosystems.com> wrote:
> On Mon, Jul 20, 2009 at 9:41 AM, Jack O'Quin<jack.oquin at gmail.com> wrote:
>
>> Despite the fact that negative nice values are ineffective for
>> achieving solid realtime audio, I doubt we'll see many distributions
>> jumping into the role of discouraging that style of programming.
>>
>> Most distribution developers see their role as packaging Linux
>> applications in a form that makes them easily accessible to end users.
>>  They generally avoid highly technical discussions about "how those
>> applications should be written".
>>
>> If enough users want to run "nice-audio" applications, they are likely
>> to enable that behavior.  Why shouldn't they?
>
> given that many distributions have actively resisted enabling the
> correct approach to writing such applications, i don't see why they
> should not be encouraged to reverse themselves on both fronts: enable
> the right way, and discourage the wrong way. it is crazy to claim that
> they simply want to make things easily accessible to end users - the
> debian packagers, for example, have argued that using SCHED_{FIFO,RR}
> is wrong and that no app should be using memlock. so, they *do* take
> positions ... i'm just saying they need a new one, and that is that
> making lower nice values available for *this* purpose is wrong. there
> may, of course, be other reasons to permit it.

I see your point and share your frustration.

To be fair (not always easy), their objection to genuine realtime
scheduling rests on a perception (real or imagined) that these tools
may open up Denial of Service attacks on *all* users -- not just the
few doing serious audio production.  Although I am not aware of any
genuine DoS attacks of that sort, distribution developers still have a
responsibility to take the possibility seriously.

So even here, I believe their motive remains to make things as easy as
possible for (most) users to do (most) everything they like.  For a
general-purpose distribution, that's probably the best anyone can do.

>From their point of view, "nice -10" is quite harmless, so they see no
need to disallow it.
-- 
 joq