[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: [wli@holomorphy.com: Re: NMU: kernel]



On Mon, May 24, 2004 at 03:18:07PM +0200, Christoph Hellwig wrote:
> On Mon, May 24, 2004 at 03:02:46PM +0200, Sven Luther wrote:
> > > Anything else is a maintainance nightmare in the long term.
> > 
> > Sure, but it provides for localized testing before large scale
> > deployement later on.
> 
> Shouldn't a distribution kernel by production and not testing of random
> changes?

Which is why we have a (rather long) development and testing phase,
don't we ? That is what dbeian/unstabvle is for. 
> 
> BTW, did you review asfs and the orinco patch properly before including?

No, naturally not, Well, the orinoco patch is from Jens, so i will have
him respond on this, but for the asfs driver, i had contact with both
the author of the patch and the author of the SFS filesystem, as well as
the whole SFS user community. Still i included it, in the current state,
because i know damn well that if i didn't, the whole pegasos user
community would rather build their own kernel, like they do for
powermac, rather than use the prebuilt kernel, which means i don't get
feedback on other issues on this subarch. So you see, there are a lot of
considerations apart from just the pure quality issues.

Also, let me give an anecdote. When i first took over the powerpc
maintenance from Dan, i was no way a kernel expert, but nobody else
bothered so i decided to launch myself in this, i made a mistake and
compiled a kernel which would not boot on powermac. Weeks passed before
anyone even noticed, and then i got feedback from only a few persons.
Any calls for testing on powermacs at that time had mostly gone
unanswered, let alone on non pmac subarches or oldworld. It took time,
month even, to get to the point we are now, were there is a package that
is actually also used by the users, but i guess most of our feedback
still comes from the debian-installer.

> I haven't seen you as either filesystem or network driver experts on
> linux lists.  Note that this is not a personal attack, I'm no net driver
> expert either and wouldn't claim I alone could review such a driver.

Well, i am not a kernel expert. I believe i am an expert in debian
packaging though, which probably makes me today more adequate than you
to do the job. But ideally we should work together, and not try to
despise the job done by the other because it is not the way it should
be.

> I am familar with filesystem drivers thouch and asfs doesn't pass a
> review.

Information on what needs changing and even patches are welcome :)

> This is one more reason I'd like to see a maintaince team instead of
> gazillions of arch maintainers in their own small tower and let as
> much work as possible upstream.

Well, as you see, the powerpc package is now maintained by a small team,
namely Jens and me.

> The kernel is a very complex piece of software, and there's very few if
> any people understanding all parts good enough to make educated decisions
> whether a change is okay or not.

Yeah, but i believe it is not only a question of understanding the
kernel, but there is also a large part involved which consist in
understanding the userbase, and our relationship to those, and who will
use the prepackaged kernels, and who will build their own stuff anyway.

Friendly,

Sven Luther



Reply to: