[Wikipedia-l] Apology

[Magnus Manske]

> Or, indeed, beyond.  Which is why you don't (or at least
> *shouldn't*) roll out
> new code into production use without stress testing it.

It was out for month on the sourceforge server, and for a few weeks on the
bomis server (as test.wikipedia.com).

> I appreciate that there is an old joke that goes:
>         "Fast, cheap, correct - Pick any two"
> but this is taking the piss.

Well, this was quite helpful.

> IM-not-so-HO, at the moment, the 'pedia  is quite useless.
> (I've been searching for the UN convention on human rights for 10 minutes.
>  I know its there, I added it in myself.
>  Last weeks wiki would've found it in 5 seconds
>  Brittanica.com just found it in 5 seconds).

The code itself performs quite well. On my own machine, it is faster than
the same function on the UseModWiki via DSL.
The performance problems come from some unidentified bug in the
apache-PHP-MySQL triangle. Currently, I'd bet on a buggy apache server
(there are some bad versions out there).
You probably noticed that sometimes, the software is quite fast and usable,
while two minutes later, it becomes awfully slow. That's the problem. It
should be solved within a few days max.

> PS: The search has returned now.  It says this
> ------------------------------------------------------------------
> -----------
> Wikipedia articles
>  Found 76 occurences of un convention on human rights. For each
> article, you
>  can see its first paragraph and the first paragraph that contains ''un
>  convention on human rights''
> ------------------------------------------------------------------
> -----------

Strange. Never did this to me. Actually, I ran your search right now. It
showed me the results (which were not very useful, but all contained the
query words). Took about 20 secs. Takes about 3 secs on my local machine.

When the software repository goes online (soon, hoepfully), be my (or
Jimbos) guest to "fix" the code...

Magnus