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

Re: [Sheflug] RAID and kernel upgrading.




On Thu, 8 Jun 2000, Will Newton wrote:

> l-k does not require subscription to post, and there are web interfaces to
> the archives.

No, but to post to the list & then trawl the archives to track the query
you may have well subscribed temporarily, it makes no difference. There's
still a mountain of crap to search through.

> > The most competitive areas are the bits of the pie most valuable: i.e.,
> > growing rapidly, most sought after, most visible, etc. Places without
> > competition are generally things people aren't interested in, hence no
> > money..
> Not necessarily. e.g. Lineo don't seem to have too much competition (OK,
> so they bought a lot of it)

I think Lineo are a bit of a non-example: okay, they buck the trend
somewhat, but what I said is true in the general case.

> > To a great extent it depends what you define as 'kernel'. Certainly,
> > drivers make up a heck of a lot of the code, they run in ring 0 / kernel
> > space, but most people don't think of them as kernel. The core stuff, MM,
> > VFS, etc., is only looked after by a few people, but then, this isn't
> > usually the stuff that's breaking.
> From about 2.3.51 onwards the VM was quite badly broken. In the end Juan
> Quintela and Rik van Riel fixed it. Rik is the guy who is "in charge" of
> #kernelnewbies, trying to get more people into kernel development.

He was also the guy who broke it in the first place.. ;)

> > That kind of then defeats the 'many eyes' principle also, surely? As many
> > people as possible should be running the dev kernels, on many
> > architectures, otherwise you end up with broken stable releases.
> They are not to be used in distros or by people who don't want to risk
> crashing production systems. If you have the time, inclination and
> expertise you can use dev. kernels. The tag of "unstable" is a warning for
> the unwary.

Yes, and I don't disagree with that. BUT, you can't say that people
shouldn't be running these kernels, because they absolutely should.
There's always a few boners, the really completely broken ones, but
they're found out within a day or two usually. There are dev releases
sufficiently stable for people to use properly. If people don't use them,
they can't be tested. Indeed, not enough people use them currently, which
is one reason why 2.2 was so badly screwed when it came out.

> > If it's labelled 'experimental', and hooks nowhere else, I don't see what
> > the problem is?
> There is a suprising amount of new stuff inside 2.3. The SMP stuff for
> example has been much advanced, which does affect many areas of the
> kernel.

We were talking about features added in deep freeze, not development!

> > I think there was also talk of remodelling part of the PCI spec to be more
> > like USB.. USB also requires devfs I believe. Ho hum...
> No.
> I don't think anything REQUIRES devfs at this moment. IMHO it should have
> gone into 2.5, but Linus seems to like it.

Not strictly requires, no, but doesn't make a whole lot of sense without
it. /dev/ is fixed devices, USB, new PCI, et al, are all hot-swap. Even
PCMCIA, to an extent. /dev doesn't deal with these at all well, and to be
honest /dev has had it's chips for a while. It's just a rubbish system.
Can you imagine what /proc would be like, instantiated statically? 

Cheers,

Alex.


---------------------------------------------------------------------
Sheffield Linux User's Group - http://www.sheflug.co.uk
To unsubscribe from this list send mail to
- <sheflug-request [at] vuw.ac.nz> - with the word 
 "unsubscribe" in the body of the message. 

  GNU the choice of a complete generation.