[BSDcert] Initial thoughts

[Tillman Hodgson]

On Sun, Dec 19, 2004 at 12:43:43PM -0500, G. Rosamond wrote:
> >I don't want to tie it to a single organization, the idea was more
> >like a sign (like those at the carnival) that says "you must be -->
> >this tall to play here". Provide a list of the basic Unix certs (and
> >there's a lot of them!) and require sometime attempting the BSD cert
> >to hold one of them (any of them, so to speak).
> 
> The reality is many BSD people are anti-cert, so I don't know what
> we're piggy-backing off of.

I don't see the anti-cert folks as relevant -- they wouldn't be
interested in a BSD cert either.

I'm simply saying that rather than try to certify *all* of BSD, let's
just certify the BSD-specific topics. Other certs handle the basic stuff
already, there's no need (or real incentive) to duplicate that and
there's many downsides to trying.

> >Since it covers only basic Unix stuff, I don't think the danger to
> >our model is very high. We also don't need the involvement of the
> >other certifying organizations -- the student just needs to hold a
> >cert from *somewhere*, that's the extent of the involvement.
> 
> To me, it's all about easily justifying a BSD Certs own existence. . .
> this undercuts it severely.

I don't see how. If I hold, say, an AIX Basic Admin cert a hiring
manager could perhaps extrapolate that I know Unix in general to a
certain level. But he couldn't extrapolate from the AIX cert that I
understand BSD things like RCng or or the sysctl system or UFS2
snapshots or [...]. The BSD cert I would hypothetically have tells him
that, and that's the value it provides.

Bundling-in basic Unix stuff doesn't seem to add value to a BSD cert.

> >It gets us out of the Yet Another Basic Unix Admin Cert game, where
> >the path is rocky and the list of failed certifying organizations is
> >long.
> 
> What do you mean "the list of failed. . "?  I'm a bit confused.  Oh,
> you mean certs that start then cease?

Brainbench, for example, seems to a null entity in the minds of the HR
folks as far as Unix certs go. It still exists, but I don't think
anybody takes it seriously (assuming they even know what it is).

I would hazard a guess that vendor certs have the advantage of being
much easier for a corporation to pay for ("We just bought an X, let's
get Bob certified in X from the vendor of X") than more open or
community oriented certs. Yet, in the open source world, the playing
field seems to have boiled down to LPI and RHCE (of which only one is
vendor based). The other vender certs seemed to have died on the vine. I
think that's because mind-share of hiring managers is the primary goal:
only a near-monopoly of mind-share can really be considered
"successful".

So we can try and compete with that, or we can chose a different playing
field. I think being the best BSD cert is achievable ;-). I don't think
being being a widely recognized Unix cert that has BSD content is
achievable. A minor difference in focus perhaps, but I think it changes
the game significantly. If we're not going to compete on the basic Unix
cert stuff, why even include that? *Poof*, a bunch of work that doesn't
help us achieve our goals disappears. Most excellent :-)

This might be a good place for me to link to the SAGE (System
Administrators Guild) annual salary survey. It lists the most popular
certs by count of those holding them (and participating in the extensive
survey). Anyway, there's some data worth mining there. The 2002 survey,
the newest for which public download is allowed (for those that aren't
SAGE members) is at http://www.sage.org/salsurv/).

> That's why the foundations are important.  And the backing of each of
> the projects.

True, from the political side of things. The foundations might give us a
way to have _someone_ sign partnership agreements. I think that'll be
important once the basic groundwork is done.

I'm not sure the projects can pay for things like psychometric
evaluation of questions, consistency testing, study materials, web
hosting (maybe this one they can), advertising, etc.

> It would make sense to do a comparison of various relevant certs in
> terms of form, eg, exams required, number of questions, format, etc.
> More for the wiki. . .

Definitely worth doing, I agree.

> >I like your idea of splitting the OS from applications. The OS we
> >have more direct control over (through the various BSD projects), so
> >the tests would need to be updated less often -- and when it would
> >need to be updated would be clearer to all of us.
> 
> Very true. . . but there should at least be sections on www, mail,
> dns, etc, configs.  These exams are aimed, IMHO, at sysadmins, not
> developers.

Right. I wasn't thinking of things like APIs, I was thinking of things
like RCng, /etc/ttys versus inittab on other Unix variants, etc.  Those
daemons that are "bundled" (/usr/src/contrib/) should probably be
covered.

-T


-- 
"The Web is the sum of all human knowledge plus porn."
    -- Ron Gilbert