Re: buffer cache patches - anybody willing to summarize?

Harald Koenig (
Tue, 5 Aug 1997 09:30:08 +0200

On Aug 04, Jeff Wiegley wrote:

> > drivers/net/tulip.c needs an update to support a number of recent cards/chips.
> NO! I have to say I agree with Linus...

I'm happy to see that Linus doesn't agree with you on this topic :^)
as he just wrote about his first pre-2.0.31 patch:

Currently the pre-patches have there larger patches:

- updated ide, aic7xxx and tulip.c drivers

> if people *really* need a driver update they can do one of two things:
> 1) ask the driver maintainer for the latest stable driver patch against
> 2.0.31 when it is released, or...

this is not an option if you're starting to use Linux, buying some CDs
and not having a internet connection and not the necessary skills yet
to ask the right and important questions (but only "why doesn't Linux work for me").
that way, world domination won't work...

> 2) go purchase hardware that is known to be supported.

DEC 21140 chips a known to work for a long time now (I think those have been
the first 100Mbit cards being supported, right?) but some recent cards
and chip releases need newer drivers. but our chips are still labeled
DEC 21140, so how should I know (and I bought those cards as replacements for
3COM 3c905 cards exactly *because* I knew they should work fine...).

similar seems to be true for AHA29xx (no idea, no longer using Adaptec products
for a long time, and now 3COM is on the same list for me...)

> I almost don't think driver updates are even appropriate at all for the
> 2.0 tree. Security and fatal bug fixes *only*. other than that 2.0
> supports what it supports and you should wait for 2.2/3.0 to include new
> hardware support.

that's not realistic and Linus did state the growth of 2.0 differently
when it was released. he said that debugged, tested, stable drivers (even new ones?!)
might be migrated from 2.1.x to 2.0.x.

otherways, we'd have to reduces the stable revision release cycles to ~3 months
or Linux just won't work for many newbees...

> adding new support and features to 2.0 just gets us back into more bugs
> to fix for 2.0. One fo the great things about linux is that 2.0 is

things that *are* debugged and tested by a wider range of users should be
able to make it into 2.0, but only then!!
I'm not talking about the 2.0.29->2.0.30 changes here;)


All SCSI disks will from now on                     ___       _____
be required to send an email notice                0--,|    /OOOOOOO\
24 hours prior to complete hardware failure!      <_/  /  /OOOOOOOOOOO\
                                                    \  \/OOOOOOOOOOOOOOO\
                                                      \ OOOOOOOOOOOOOOOOO|//
Harald Koenig,                                         \/\/\/\/\/\/\/\/\/
Inst.f.Theoret.Astrophysik                              //  /     \\  \                     ^^^^^       ^^^^^