Re: MJR on Linux/OSS

[Darren Reed <darrenr () reed wattle id au>]

My apologies for not reading your article, I just couldn't believe
that you'd have said what someone was claiming you did.

> Nope. My argument is that it's becoming frighteningly
> difficult to build code that you can reliably install and run
> across Linux distros, let alone Linux/BSD distros.

This is the late 1980s, early 1990s all over again where
there were a dozen Unixes and all slightly different.

Of course many of those who are now responsible for all of
the distros having their own personalities never had to
deal with that and so weren't there to experience *that*
mistake.

Posix, etc, went some want to fixing this but unless you want
to stick to using a small selection of library functions (and
forget gtk), Posix is almost irrelevant.

> The
> difficulty in building products (not user interfaces) that
> can reliably install cleanly makes it difficult for a vendor
> to support all the distros.

It's not just APIs, either.  Various commands have different
quirks on different distros, etc.  It's simply ludicrous.

It's frustrating to have to deal with.  I can get IPFilter
working on *a* distro and if I'm lucky the kernel source
will be mostly the same on another, never mind the kernel
source provided by the distro often not being in a form
that builds a new kernel, but I'll still have to write a
custom startup script.

> Which pushes them towards
> "appliance-izing" the OSS kernels, so they can re-achieve
> a necessary level of control. If you're a vendor selling a
> product on *NIX nowadays you are completely hosed if
> you don't QA on each distro. Which means you need
> a whole room full of waste-of-time. AND you still get
> hosed because some customer uses whatever-server
> instead of blah-server and everything blows up. So by
> being so diverse, the OSS *NIX variants are dooming
> themselves to become embedded appliances or
> evolutionary dead ends.

Have you got the warehouse of slaves in India, yet ?
Because those rooms are going to be way too expensive,
if you want a competitive product, if you fill them in
any 'western' country.

This is the hidden cost of using OSS to build your
product - you now need to hire a kernel hacker, someone
else to hack on userland, a test engineer, etc, because
your applications people don't have the skills or time
to waste on it and because of your customisations, you
now need to merge all the patches yourself.

> What really makes me shake my head in dismay is
> the short-sighted stupidity of the OSS community as
> it now tries to re-converge on a "standard linux distro"
> as if it wasn't ONCE A SINGLE STANDARD DISTRO.

It won't happen.  There's too much ego at stake.

> I remember when there was one BSD and one
> LINUX and it wasn't that long ago. The OSS community
> should have "put all the wood behind one arrow" and
> made that one distro as good as it could possibly
> be, instead of making 200 distros, each of which
> has some great stuff, and some braindamage, and a
> small army of fanatical adherents who throw rocks
> at the other distros on mailing lists. It's pathetic.

That's what happens when you have people who can't agree
on how XYZ should be done and rather than compromise, they
just go off and start their own project with some group
of ad-hoc people.

This is the dark side of OSS.

Darren
_______________________________________________
firewall-wizards mailing list
firewall-wizards () honor icsalabs com
http://honor.icsalabs.com/mailman/listinfo/firewall-wizards