Jobs...

[Vlad]

On 7/20/06, Evan Leibovitch <evan-ieNeDk6JonTYtjvyW6yDsg at public.gmane.org> wrote:
> Vlad wrote:
>
[snip]
>
> > Just because something is cheap, or free, and it mostly  "works",
> > doesn't mean that it should be used.
>
> I'm sorry, but that statement just reeks of elitism and has no relevance
> in today's world, and I'm not just talking IT. In every field there are
> elite products that purists love and the rich/enthusiasts can afford.
> Sometimes, but rarely, the elite product is the most cost-effective
> solution to a need. The rest of us generally choose products that by
> comparison are "cheap and mostly work", that dramatically outsells the
> elite stuff. Anyone who buys PC-brand food, takes generic drugs, flies
> US-based airlines, watches analog, broadcast TV, drinks tap water or
> drives a Suzuki knows that. Why should IT be any different?

        Perhaps I should expand my point a bit more (though I do admit
that it is elitist).
        Most of my gripes with Linux people (management, clients)
wanting it to do more than it is currently designed to do. I'll try
and avoid the mid-tier gray area, and stick to the higher end, such as
high-end SMP, clustering, HA, et al.
        I've found that certain businesses take Linux to be their holy
grail, and order their denizens to just "make it work". Which,
usually, means a hack on top of a hack, to just get it out the door
working. There's lots of good software, even for the enterprise. But
not so much that's "great" or "excellent".
        I'll argue that there's a lot more elitism coming from Linus
about keeping everything GPL Open Sourced and deliberately breaking
Kernel internals to irritate Closed Source driver developers, than in
my statement. This isn't just about the regular changes to tick off
nVidia. I mean having to spend weeks troubleshooting NetBackup's SCSI
tape device interaction on 2.6 Kernels. Or a craptastic 2.6 (earlier
2.6 versions) scheduler. This has now been fixed, but random overhauls
in a stable branch are not so good. (RedHat & Co. have to devote many
resources to keeping a "stable" and consistent Kernel tree for their
distros; it's a Herculean effort, really.)

        I want a -Server Kernel tree, and a -User Kernel tree. Not
Kernel 2.6.9-34.ELsmp-I_lost_track_of_my_rebuilds_to_make_it_work.

        I'll take the Sun guarantee that Solaris 8 binaries will run
without any changes needed, or with new behavioural side-effects, on
Solaris 9, 10, etc.
        That can't be said about 2.4 to 2.6, or even in between
2.4/2.4, or 2.6/2.6. (Think the MM overhaul in 2.4.9(?), and the
scheduler changes in 2.6.something.)

        Just my two cents. And a rant because management in some
places I've been Just Wants It To Be Free(tm), and for their staff to
Just Make It Work(tm).


        As a side note, there's also some hilarity in Linux-land with
high-uptime and application interactions. For example, sar breaks on
RHEL4 after 147 days or somesuch. Just completely breaks, stops
logging, and reverses the idle/usage fields. There's other
applications that can't go above 300something days, and just die hard.
It's amusing to work in a place where most Linux boxes get 300-500
days of uptime. I miss Solaris.

> - Evan
>
> --
> The Toronto Linux Users Group.      Meetings: http://tlug.ss.org
> TLUG requests: Linux topics, No HTML, wrap text below 80 columns
> How to UNSUBSCRIBE: http://tlug.ss.org/subscribe.shtml
>


-- 
end
--
The Toronto Linux Users Group.      Meetings: http://tlug.ss.org
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://tlug.ss.org/subscribe.shtml