Tino Wildenhain wrote:
>
> I think the chances here are to make at least one postgres based
> certification which is really about sane database design - e.g.
> can be applied on many other rdbms. This includes detection
> of steps to improve queries by reading the explain output for example.
>
> This means you need a practizing kind of process and not the
> simple multiple choice (most test vendors go practice tests
> these days). The other problem is the international trustworthy
> hosting of these tests.
>
If I were trying to design a certification, I would do so in the
following way:
Section 1 is multiple choice, say 100 questions, and one must answer 80
of them right to qualify for the second stage. This is just a screening
stage. The idea is simply that you want to weed out the people who
don't know the basics from even trying on the more expensive/labor
intensive stage.
Section 2 would give you a complicated specification and maybe a blank
database or a tool like DIA and ask you to design a database to the
specification. This is then graded perhaps by the company which
designed the certification and the candidate is either awarded or denied
the certification on this basis. I would have the questions change
frequently here and try to make the specification complex enough to
prevent simple memorization.
> So first we would need an agreeement among a group of individuals,
> forming an organisation which is the authority of the certs and
> watches the fulfillment of all requirements and can actually revoke
> certifications.
>
Sure, and we can call it the X509 Certification Authority :-) Just
kidding....
Really the best certification IMO is actually having a tangible open
source project one has built. But that is not from the corporate
entity's perspective.
Best Wishes,
Chris Travers
Metatron Technology Consulting