On Thu, 8 Jul 1999, Jared Buckley wrote:
> "A.R. (Tom) Peters" wrote:
> >
> > Which brings up the question, how and when do we upgrade the exams and
> > what do we do with the old ones?
> 3) When we revise, do we completely rewrite the exams or
> merely update the relevant sections?
>
> On the first topic, I don't believe that we should be tying exams
> revision to any vendors upgrade cycle. (The exception(s) being the
> vendor specific T2 exams... Of course that raises another question of
> how often do we revise those exams?) I'd also say that we only revise
> on "production" releases, and then only those releases that are really
> significant to the daily operations of a cert holder. i.e. A release
> that revamps the SMP capabilities, but leaves the API's and user
> interfaces unchanged isn't probably a good enough reason to change any
> of the low level tests.
I think that the two criteria should be (1) exam security and (2) change
in content. I can tell you that the books I bought a year or 18 months
ago (which were written six to 18 months previously) are definitely stale.
I think Linux is currently changing sufficiently in a year or two to make
new exams a good idea.
Definitely, the vendor-specific exams need to be fairly up-to-date. I
think MS, to name a well-known rival program, comes out within a month or
two. For big deals like Windows 2000, I bet they're quicker.
Considering security of the exam, I think an ideal would be revision every
six to twelve months. It would be OK (perhaps preferable) to have item
overlap. (But I don't think that LPI is nearly ready to keep up with that
ideal pace.)
If you don't think security is a concern, consider that training outfits
send people into US educational exams to memorize items. Then they write
a carbon-copy and teach the test almost exactly. This decreases the
exam's validity. The largest US testing organization, ETS, had to
withdraw it's national college exams a couple years ago because someone
compromised enough of the entire pool to call their exam's validity into
question. NOTHING ruins your day as a test publisher faster than someone
destroying the validity of your painfully developed instrument by making
your trade secrets (i.e., the exam items) public.
You could always disclose old items in practice exams. This is
an excellent practice which will increase the exam's validity by
decreasing the novelty of test taking (allowing true competence
differences between test-takers to shine though more brightly) and the
Linux communities' value from the LPI exams.
> retiring them and replacing them with the new exams. (A question to
> carry over into the recertification discussion would be: do we make old
> cert holders retake the new exams?)
The new and old exams should be made to be roughly psychometrically
equivalent so that passing the old one indicates you would pass the new
(assuming identical content). So the only reason for retesting is
recertification (i.e., the original thread).
-Alan
________________________________________________________________________
This message was sent by the linux-cert-program mailing list. To unsubscribe:
echo unsubscribe | mail -s '' [EMAIL PROTECTED]
