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

Re: This was posted on Digg.com -- Linux Distribution blog/article

On Sat, 2005-12-10 at 06:43 -0500, Bryan J. Smith wrote:
> WTF?
> Who the freak gives you the right to assert how well I do or don't test!
> Really?  That is the most arrogant, insulting and assuming statement
> I've ever been presented with!

Now we're getting somewhere. :-)

Note that I didn't make any statements about whether your testing
practices are shoddy, just that they're inadequate.  Nor did I say that
mine were any better.  They may be worse.  In fact, just from what I've
read from you so far, I expect your testing has some nuances I could
learn from.

The point is not to slam you, but to tell you that the distro business
is hard.  Way hard.  A lot harder than you seem to realize.

You don't think you're in the distro business, because you give people
black boxes and not installers or hardware detection systems.  I get
that you don't care about those things.  But you do ship bits, don't
you?  Those bits look a lot like a Linux system, right?  What do you
call that?  And, more importantly, how do you keep that working and
deliver new features?

That's a distro.  Anaconda and Kudzu are not a distro.  A working Linux
kernel and a storage system with applications: that's a distro.  And you
have to maintain those bits somehow.  How do you do that?  That's what
we're talking about here.

You seem to think that this job is somehow split into two groups, as if
you're exempt from the kinds of things Red Hat has to do.  Sure, you're
exempt from some of them: making a pretty installer, for one.  But
you're not exempt from others.  And what I'm trying to tell you is that
you're throwing away a lot of good work when you insist that you have to
build everything, work Red Hat has done and that you should do too, but
can't because you don't have Intel and HP and Dell sending you free
computers to test your little distro on.

Because when you build a new Gentoo system, you're combining a lot of
pieces in a way they've rarely, if ever, been combined before.
Probably, they will all work fine, especially if you test it.  But
you're testing a brand-new environment in a very narrow range of
possibilities.  What kind of assurance can you give your customers that
your platform hasn't developed a brittleness that won't show up until
circumstances change slightly at some unexpected time?  You can't.  No
single person can.  But Red Hat and Debian have a decent shot at it.
Why not pass that decent shot on to your customer, as well as you can?

> You should consider that maybe, just maybe, someone like myself has been
> working with GNU/Linux for 12 years in industries like ... say,
> aerospace and semiconductor design?

[shrug]  Maybe you should consider that you're talking to someone who
may be just as experienced, or even more so.

I don't like to play the ad hominem wars, but if you want to bring it,
fine.  I have two more years of experience in Linux than you do.  I ran
Linux when it couldn't boot multi-user, and ran other Unixes before
that.  I've been a Debian developer for close to ten years, and a
Progeny employee for five.  I was the lead engineer for Progeny Debian
1.0, a distro that won a lot of nice reviews when it was released.  I've
written a Debian installer from scratch, and enhanced at least two
others.  At Progeny, I've participated in the building of black-box
distributions as well as consumer-level ones.

More to the point, I make my living cleaning up messes created by people
who think they can fully test a distro all by themselves.

But really, all this is just posturing.  What matters is what we know,
and how well we know it.  I think I know the distro market, but who
knows?  Maybe I'm a fool, and all those happy customers of mine are
headed for a fall you can save them from.  Maybe I can save them, if you
can teach me before it's too late.

But so far, the only example you can bring to the table is rebuilding
the Apache stack, and you express incredulity that I can do this easily,
in package format, without all this ports infrastructure.  OK, great.
Not everyone can be an expert.  But now I'm supposed to be impressed
that Gentoo is so great that it lets people hack the Apache stack, which
is "so much harder" in packaged distros?

What that tells me is that you don't know how to do those things in a
package distro.  Did you know that Debian delivers source, too?  Did you
know that you can download the source for any package with a single
command, build it with a second command, and install it with a third?
Did you know that Debian has people who rebuild every single package
every so often, looking for bugs, and that this happens automatically?
Did you know that only about 10% of the binary packages Debian ships are
built by hand, and that all the infrastructure they use is also
available for you to run yourself?

Maybe you didn't.  That's OK, because I don't know how all that stuff
works in Gentoo.  (And I still don't, by the way.)

But anyway, what I'm looking for from you is something hard, and how
Gentoo makes it easier.  Here's the level I'm looking for: hack the
Linux dynamic linker to support layered ABIs for LSB compliance, make it
work side-by-side with the standard dynamic linker, and deliver it on
three architectures: standard Intel x86, AMD x86_64, and Itanium.  That
was my last big project.  That's what a hard problem in distro
maintenance looks like.

Frankly, I shudder at the thought of doing something like that on
Gentoo.  I have tools that track things for me in Debian: complete
control over changes, easy multi-arch rebuilding, easy replacement and
rollback if I screw up, etc.  But maybe you can tell me that I'm wrong,
that Gentoo has all those things, and even more that just rocks.  Should
I get pissed and end the conversation because you're telling me
something that contradicts my cherished beliefs?

So maybe you should consider that you're in my area of expertise, and
that maybe you can learn something from me.  The reason I started this
conversation was because I was interested in learning something from
you, in your area of expertise.  Because, despite all the C.V. crap I
posted above, I know there's a lot I don't know about, and Gentoo is one
area where I'm really clueless, and you sounded like you had a clue.

But if you're not interested in having your knowledge questioned, then
oh well.  I suppose I wasted my time.

To unsubscribe, send email to majordomo@luci.org with
"unsubscribe luci-discuss" in the body.