On Thu, Nov 09, 2000 at 10:33:39AM +0000, Dag Brattli wrote:
> Hi,
>
> Jean and I have been discussing with Linus about why he won't accept our
> patches for Linux-2.4. Basically hi won't apply the patches because they are
> large and ugly ;-) but Russell King has already shown that small well documented
> high quality IrDA patches will not make it into the kernel either. I guess Linus has
> something against IrDA (even if Transmeta is a member of IrDA these days). Now
> he's talking crap like us having to study the art of submitting patches so they
> will eventually get accepted by him (since his packet loss is so high)
Dag, I would not go as far as that. Don't put the blame
entirely on Linus, some of the things he said is valid. A cc on the
kernel mailing list and a detailed changelog are nice touches that can
make him happy.
On the other hand, it is true that getting a patch in the
kernel can be more work than doing the patch himself.
> Who wins the fight between Linus and God?
>
> Trick question! Linus _is_ God!
The point was not to win the fight. The point was getting
Linus attention and making him recognise that something is broken in
the current process.
6 month ago I sent a similar e-mail about Pcmcia, when the
code in the kernel was unusable (it is still unusable for me). Now, it
was about IrDA. There was ISDN, there was Don Becker network
drivers... The hope is to establish a pattern that make him recognise
that he could make life a bit easier on the maintainers by giving more
feedback.
Don't worry, you were not comfortable during this flame, but
I'm sure he was not either. Wait until we see the comments on kernel
traffic ;-)
> Coding for Linux-IrDA is something I do because it's fun, and being flamed by
> Linus is not my idea of having fun. What a way to treat people helping for
> free. Makes me want to switch to FreeBSD since they at least have their elected
> core team of kernel developers who decide on such issues. But I probably don't
> have time to port the stack to FreeBSD, and since I don't have much time for
> Linux-IrDA anymore either (not full time anyway), I want to make sure all the
> unpaid time I use for Linux development is 100% fun! If not I'll do something else.
"The grass is always greener in the field next door". I'm sure
FreeBSD management has its share of problems. For me, there are too
many things missing (Wireless LANs) to be productive.
> I'm not saying that Linus should accept every patch, but I'm not
> so impressed with his skills as a manager. One thing is for sure: I don't have time
> to split our current 300K patch into smaller patches with higher acceptance factor.
> It should be good enough, that the maintainer of the IrDA subsystem has blessed
> the current patch (at least when it doesn't touch any other files than those in the
> irda directories of the kernel, and when everybody knows that the current
> irda code in 2.4 isn't working anyway).
I agree, I don't fell like splitting the 300K either.
You have a valid point there, the point of having subsystem
maintainers is to delegate responsability and reduce the load on the
top. As Linus doesn't read the mailing list and doesn't do proper
testing, his filtering of the patch will always be superficial.
> I think we would be better off maintaining the stuff ourselves, and the advantages
> for being inside the kernel distribution are very small. We end up getting
> our patches rejected, and we cannot fix a bug right away because we have to wait
> for the next kernel-release (and then we cannot be sure the patch have been
> accepted, and kernels are released where the IrDA subsystem doesn't work). I
> remember somebody saying that it was more fun in the old days of Linux-IrDA
> (before we got into the kernel distribution), and I feel the same thing myself. If
>such
> a thing happened then we will go back to do the same thing as with PCMCIA
> (before 2.4). People will have to download Linux-IrDA and do a "make
> install" (modules only), but distributions like Mandrake etc should/would have it
> pre-compiled and pre-installed. Should not make things much more difficult for
> distributions, since the irda-utils would always match the modules inside the
> package. And at least PCMCIA used to work back then ;-)
Pcmcia do work, as a standalone package. I can't use the code
in the kernel. There again, the reason is that Linus did bypass
totally David Hinds and started hacking like crazy the Pcmcia code in
the kernel. David was not feeling good about this and decided to
ignore the code in the kernel.
For now, let's keep the status-quo. Let's keep Linux-IrDA as a
big patch against the kernel. Applying a patch is not much more tricky
than compiling a standalone package, and it minimse the work you have
to do. Do a small do explaining how to get IrDA up and running.
Then, I expect Alan to take over the maintainance of 2.4 when
Linus will for 2.5. Alan has always done a superb job with respect to
kernel maintainance, and has always been receptive to your
patches. So, at this point, you create a nice diff, with changelog,
send it to the kernel mailing list and ...
> So should we make our own Linux-IrDA package, or should we stay within the
> kernel distribution? Currently we don't need any modifications within the Linux
> kernel itself.
Keep the current process for now. Just post regular patches on
SourceForge. By the way, why the latest patch has not been uploaded on
source forge ?
> Any comments?
>
> -- Dag
Have fun...
Jean
_______________________________________________
Linux-IrDA mailing list - [EMAIL PROTECTED]
http://www.pasta.cs.UiT.No/mailman/listinfo/linux-irda
