[Ksummit-2012-discuss] [ATTEND] Your upstream maintainer just isn't that into you...

Dan Williams dan.j.williams at intel.com
Wed Jun 27 20:44:58 UTC 2012


On Tue, Jun 26, 2012 at 2:58 AM, Mark Brown <broonie at sirena.org.uk> wrote:
> On Mon, Jun 25, 2012 at 06:44:56PM -0700, Dan Williams wrote:
>
>> I authored the article "Avoiding the OS abstraction trap" [1], which
>> raised some points about the platform problem [2].  I think the
>> platform problem also contributes to these occasions of unbounded
>> latency getting patches upstream.  It is simply much easier to merge
>> self-contained code and this becomes another unfortunate incentive to
>> avoid touching the "platform".
>
> I know one of the most common reasons I push back on things is that
> people either aren't using existing subsystem features or should be
> adding new ones.
>
> I think this depends a lot on how much of the work is factored out into
> the subsystem.  Once you've got a lot of code factored out into the
> subsystem I find the balance goes very much the other way and it becomes
> much easier to have code in the subsystem than in the drivers - it
> reduces minefields for refactoring so much if you don't have to worry
> about special cases and well written drivers start to look much more
> dumb.

I think that's the definition of a right-sized maintenance load, one
where the maintainer (and sub-maintainers) have the bandwidth to spot
the inbound patches that are avoiding the platform.


More information about the Ksummit-2012-discuss mailing list