[OpenAFS] OpenAFS on Linux 2.5.x

jon+openafs@silicide.dk jon+openafs@silicide.dk
Thu, 17 Apr 2003 17:31:50 +0200


On Thu, Apr 17, 2003 at 09:14:45AM -0400, Derrick J Brashear wrote:
> On Thu, 17 Apr 2003 jon+openafs@silicide.dk wrote:

[cut]

> > > You missed the importance of "50 variants of what purport to be the same
> > > version of the Linux kernel."
> >
> > Just make it for the default linus kernel then. It shouldnt be hard
> > to avoid conflicts with other stuff, or doubling their work.
> 
> have you missed the constant requests for redhat 9 support? for support
> for updates of older redhat versions? those aren't unreasonable requests,
> but they're not requests to support 2.4.19, they're requests to support
> something that's 2.4 with new features from 2.5 included.

Yes, and for that i would say... Use a default linus kernel.


> supporting kernels which vendors distribute constitutes a large portion of
> the development time spent on linux, but it makes sense to spend available
> resources on this as opposed to 2.5 support, because the kernels are out
> there now, being used by large sets of people, at home and at companies,
> who want to use afs.

These people could just as well be using default 2.4 from linus
(yeah yeah marcello? but linus gave him the 2.4 tree)
I think that trying to catch 50 moving targets is much more
work, rather than supporting 2.4, and moving towards 2.5


> > > Right. And all I'm saying is, you can like my attitude or not, but nothing
> > > says the person who interfaces with linux kernel people has to be me;
> > > Nobody's falling over themselves to do it, though.
> >
> > no, but it would have to be someone smart :)
> 
> then clearly i'm not your man.

i think you are, but it wont work if you dont by yourself
take on the job (or unless i pay you to do so, which i cant).
So i figure someone else would have to try to work with it.
I believe that it is a little bigger than i can bite through.


> > True keeping it a patch/module that can be loaded would allow
> > us to change more stuff in the kernel to get openafs to work,
> > but is that a good thing? We might end up doing more work as
> > the internal kernel stuff might change in the future, and besides
> > i feel that being in the kernel gives extra credability, which
> > i would prefer, since then more people would use AFS.
> 
> means turnaround on adding new features will be slower.

No, i dont think so. Those that wants those features can add
a patch to the running kernel, i LVM2 uses this approach for
2.4. But i guess you could be right, i just prefer a stable
mature product rather than new features, besides, how many
new features are needed ?

Anyway, i guess i'll stop now, i dont want to take more of
your time away from openafs, it is there it is needed.




JonB