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

Red Hat and kernels [Re: ISDN with RedHat 6.0?]



On Thu, May 20, 1999 at 04:40:29PM +0300, Alex Shnitman wrote:

> this general approach -- of forking your own version instead 
> of doing the regular well-established route of submitting the patches
> to linux-kernel -- makes me sick.
> 
> What the hell is going on here? How do people accept this kind of
> thing so lightly? I don't get this...

I agree with several of your points; I think that their diffs from
the mainsteam kernel should be better marked and that it should
be possible for anyone to "redhatize" any kernel by applying this
diff set. However, permit me to make two observations:

1. Red Hat are producing a system, not bundling software. They
   are integrators who make their living by getting many things to
   work together quite well, and if you have to pay in time for
   newer kernels, you are compensated by other Good Things that RH
   do. Think of a shop that does not have a dedicated hacker that
   can tweak packages all day long. Red Hat lets them use a fair
   system that isn't bleeding edge, but still wet.   That's nice.

2. Red Hat are, IMHO, one of the best *nix vendors in terms of
   security behaviour, responsibility, etc. In fact, I'd place them
   second only to OpenBSD. They organize audits on existing code,
   are extremely prompt in fixing problems, and do not hide any
   information from their customers, whether paying or not. They
   are very professional about this.

3. (okay, I lied) Red Hat aren't the only folks splitting development
   trees. Alan Cox has been doing ac kernels since back then, and
   now Andrea Arcangeli has his own andrea series. There's also
   the def fs stuff from Richard Gooch which is not exactly up to
   sync with the main kernel (though it's not that slow to catch up).
   In short, there are many splits. So far none of them are deep
   enough to warrant BSD fear. It's a natural thing in development,
   taking "time off the make bzlilo race" to stabilize a feature or
   a set. After all, SOMEBODY has got to have nice uptimes. How are
   we ever going to achive that if we change kernels once a week? :-)

-- 
believing is seeing
gaal@forum2.org
http://www.forum2.org/gaal/