Re: Certification Available +Pronounce

On Wed, Aug 24, 2005 at 11:23:00PM -0700, Chris Travers wrote:

> As some one who holds many certifications (in order of value: MCSA,
> MCSE, Inet+, A+, Server+, Network+, LPIC-2), I feel inclined to
> chime in here...

The more, the merrier :)

> For the record, I passed the LPIC-2 and Server+ exams during their
> beta testing stage. At the end of the email I will share my
> thoughts about PostgreSQL certification, but the rest is
> certification experience in general.

> >I am not aware of any area where the opinions are "mixed" on
> >certifications, except in the sense that incompetent HR people like
> >them and the people who have to work with (or worse yet, deal with
> >the result from) certified incompetents do not.
> >
> This is a big point. When I worked at Microsoft, I was required to
> pass a certain number of Microsoft certifications per year. I
> passed the other ones to keep myself balanced and sane because I
> didn't want to be trapped working with Windows the rest of my
> life.....

> >This is precisely where we *dis*agree. Although I have met several
> >competent people who hold certifications, my experience is that in
> >the overwhelming majority of cases, a certificate tells you that
> >the person is *not* competent.
> >
> With all due respect, it depends on the certification. There can be
> well designed certifications, but these usually have a hands-on lab
> component. I passed the IIS4 (and entirely based on my Apache and
> IIS 5 experience, no less) exam without cracking a book and was
> rather amused to see it rated as one of the hardest exams in the NT4
> series. I suppose this is because it was the only exam that
> Microsoft designed that was ever worth anything.

In what ways was it "worth something?" To me, that your relevant
experience and lack of rote learning got you through the test where
cracking open their book did not would be a good sign. :)

> Similarly the LPIC-1 exams were really good. They were *really*
> difficult (but with a low passing score). But they really tested
> one's sense of fluency with the command line among other things. I
> *learned* a lot taking these exams.

That's sounding like a criterion, too.

> But this is the problem: People often see certifications as a quick
> and easy substitute for learning the technology. Of course, in the
> long run, learning the technology is far less effort than bumbling
> around a system you think you know how to use but don't really
> understand how it works, but this is not the rational most people
> have, both those who want to be technicians and those who want to
> hire technicians but don't know what they do.
>
> Finally almost all certifications end at the "technician" level. It
> is very hard to test someone's deep understanding of a technology
> without resorting to formuleic questions which are easily memorized.

I don't know that anything *cheap* can test this deep understanding at
all, even to the level of a vague guess. Writing a specification and
a rationale for same given incomplete requirements and a few
hard-to-schedule people would be one way, but it's not cheap to
arrange nor to assess.

> As far as I am concerned the only "certified engineers" are those
> with college degrees in engineering disciplines (including CS).

And even among them, there is a distressing percentage who either
spent so much time on Proprietary Product(TM) that they have no idea
what the underlying principles are, or on the other end, there are the
math-turbatory cowboys (they're always boys, however old they are) who
imagine that they can (and should) derive the entire thing from the
principles of set theory, but can't in fact code their way out of a
wet paper bag.

A good certification must protect people from both ends of this.

> To call someone a Microsoft Certified Systems Engineer is like
> saying that someone holds a Masters in Electrical Engineering from a
> Non-Accredited University (and just paid his $50 to get the
> diploma). To test this point, I had considered trying to see if I
> could pass the VB MCSD certifications without actually learning any
> real VB....

I have this awful feeling that you passed.

> The reason why I don't tout my certifications is simply that I know
> my material reasonably well and I don't want to be associated with
> those MCSE's who cannot figure out how to fix Microsoft Word when it
> opens minimized.....

And that's *just* the kind of certification that could do PostgreSQL a
lot of harm without doing it any good as a tradeoff.

> >>- experience is allways the telling-point someones ability; but
> >>when you hear that some one is CISCO certified proffessional, or
> >>Sun Certified Java programmer, or Red Hat Certified Engineer for
> >>example - a certain air of respect carries with these titles.
> >
> >What air of respect? Among people competent to make hiring
> >decisions, such a certificate conveys an air of disrespect.
> >
> Knowing what I know about the RHCE program, I would probably see it
> as a positive step. But again, all you know you are getting is
> someone you hope will be a decent technician *not* a certified
> engineer.

I know that this sounds corny, but I'd rather get somebody with that
elusive quality called common sense and a liberal arts degree.
They're likely to be able to do something that's hard to test, namely
use their imagination on a really novel situation.

> >I am not denying the possibility of a certification that really
> >means something, but that would mean that at a minimum:
> >
> >1. There would be a significant, checkable prior work requirement
> >for taking the certification exam, and
> >
> >
> This is the real problem (chicken-or-egg). Furthermore the
> definitions of work in this case would prove problematic. Where
> exactly one draws a line here is pretty tough.

Too tough to be worth pursuing? I'm thinking that an employer's
signed recommendation would be one way.

> >2. Some large percentage of those who take the exam fail it and
> >would not immediately get another chance to re-take it.
> >
> No problem here. But what do you do when vendor training is often
> offered as a part of the certification problem (like the RHCE)?

Frankly, that is a conflict of interest, and should disqualify the
certification. It is simply too hard to avoid "teaching to the test."

> >3. The exam would involve quite complicated hands-on use cases and
> >would not contain any questions whose correct answer was a quote
> >from the documents.
> >
> This is one of the things I have liked about the RHCE documentations
> is that it emphasizes hands-on work.

:)

> >These criteria are anathema to the profit motive, which is why, to my
> >knowlege, no such certification currently exists.
> >
> Aside from criteria 1, it is more an issue of degree than substance.

Is it?

> Now for PostgreSQL certification. A *real* PostgreSQL certification
> project would be extremely difficult and runs up against at least the
> following issues:
>
> 1) There is a lot of bad information out there about database design

True.

> 2) Unlike Oracle, administering the basic server is not that
> complicated. I.e. the barrier to being a technician is pretty low.

Also true. However, having a candidate be able to describe, or maybe
even demonstrate, how they'd find the answer to a question that
involves reference material would be instructive. I've been known to
ask people questions like, "what do you think of chapter 5 of book
foo" and eliminate them as a candidate if they have a detailed opinion
without consulting the book. It tells me that they are much more
likely lacking imagination and bad at setting priorities than that
they have an eidetic memory.

> 3) Very few programmers want to know how to use an RDBMS properly
> (part of why MySQL is so popular).

One perennial complaint is that programmers have a tendency to test
things with one row. Perhaps an exercise on putting together what the
person presumes to be representative sample data sets including what
assumptions were made could be a part of this.

> So you are stuck.

I think each of these things presents an opportunity :)

> Most vendor-sponsored certification programs are marketing programs
> in disguise "Look at the Cool Stuff(tm) you can do with our
> software." In the case of the MCP exams, they often fail miserably.
> In other words, they teach on features rather than substance which
> is *why* you get people working well above their ability simply
> because they have a certification (they didn't realize it but they
> were being trained to sell the software rather than use it). Do we
> really want that image for PostgreSQL?

I'm pretty sure my opinion on this is clear ;)

> If I wanted to recommend existing certifications for someone who
> wanted to get a cert that would help him be a safe DBA, I would
> suggest that he/she start with the Server+, and then take the LPI
> track. As for database design, that is another matter. Study,
> Study, Study. No certification required.

:)

Cheers,
D
--
David Fetter david(at)fetter(dot)org http://fetter.org/
phone: +1 510 893 6100 mobile: +1 415 235 3778

Remember to vote!