> > Brion Vibber wrote:
> >
> > >The code's still there (see
> > SpecialIntl.php etc), if
> > >anyone wants to work on cleaning it up and making
> > it production-ready that
> > >would be great.
That's my mess, I might have some time next week to cut it in shape ;-)
Everybody else is, of course, welcome!
Magnus
____________________________________________________________________________
Jetzt bei WEB.DE FreeMail anmelden = 1qm Regenwald schuetzen! Helfen
Sie mit! Nutzen Sie den Serien-Testsieger. http://user.web.de/Regenwald
I'd really like to get us switched over to using Larousse as the web
server soon and take some of the strain off poor Pliny.
Jason, can we get a commitment to change the DNS entry for
www.wikipedia.org over at some particular time in the next couple of
days?
I've copied in the uploaded files from the English wiki, we can rsync
the latest changes up when it's time to put it online (at which point
pliny's copy of the wiki can be put into read-only mode to avoid
confusion).
-- brion vibber (brion @ pobox.com)
> Message: 5
> Date: Mon, 12 May 2003 20:57:14 -0700 (PDT)
> From: Hunter Peress <hfastjava(a)yahoo.com>
> Subject: Re: [Wikitech-l] Kernel config
> To: wikitech-l(a)wikipedia.org
> Reply-To: wikitech-l(a)wikipedia.org
>
> reducing these options will maybe shave a couple hundred k of the size of the kernel image.
>
> triming down a kernel does not make it faster, it might save you 4 microseconds at boot time...
> There is *NO* good reason to recompile kernels unless you have to.
>
> __________________________________
> Do you Yahoo!?
> The New Yahoo! Search - Faster. Easier. Bingo.
> http://search.yahoo.com
>
Basically you are right, all except loadable module support are not
really worth the hassle.
But I think, if you want to secure things, you sould really disable
loadable module support. Most root-kits need this to work. If you can
somehow disable it, the chance of a permanent security breach are pretty
slim.
And for that to work you need to shave of those couple of 100 k's to get
all in one nice kernel package.
It all depends on how paranoid you are (or how you trust your backups
:-)
Cheers
Leo
I'm getting this error message in the German Wikipedia:
Es gab einen Syntaxfehler in der Datenbankabfrage. Das könnte eine
illegale Suchanfrage sein (siehe Wikipedia durchsuchen), oder ein
Softwarefehler. Die letzte Datenbankabfrage lautete:
SELECT cur_id,cur_namespace,cur_title,cur_text FROM cur,searchindex
WHERE cur_id=si_page AND ( (MATCH (si_title) AGAINST ('test')) ) AND
cur_namespace IN (0) LIMIT 0, 50
aus der Funktion "SearchEngine::showResults". MySQL meldete den Fehler
"1016: Can't open file: 'searchindex.MYI'. (errno: 145)".
Some time ago, there was some discussion about a new form to put the
interlanguage links in. It all looked very nice, but was not ready yet.
Then followed a long (months?) period of silence. Is anything happening
on this front at the moment?
Andre Engels
For reference, attached is the kernel configuration file for the kernel
I'm compiling right now.
It's a 2.4.20; unpatched so far as I know. (Unless Jason added anything to
it earlier.)
-- brion vibber (brion @ pobox.com)
Brion - thanks for fixing the "search" button.
I said:
> > 2. Even if you hand-fix the search, the title search doesn't work
> > at all, nor is there any hint that the search wasn't performed.
> > It should either do a title search (that's not expensive), or
> > at least say "No title search performed, server too overwhelmed".
Brion Vibber <vibber(a)aludra.usc.edu> replied:
>
> Jeebus, how many more disclaimers do I have to put into that thing?
You don't need MORE disclaimers. I think part of the problem is
that the text labelled "Results:" above the disclaimer makes it
appear that the disclaimer is irrelevant.
The current display looks like the system _did_ do a
search, with null results. It's then followed by some sort of
disclaimer, saying that sometimes searching is disabled
(during your peak hours). But since I got search results,
clearly it's not your peak hours and searching is currently
enabled. At least, that's what I thought the Wikipedia meant.
Apparantly that's not what was meant.
At least, I find it very confusing, and I suspect others will too.
When searching is disabled, I believe Wikipedia should
NOT display the following text:
"Showing below 20 results starting with #1.
View (previous 20) (next 20) (20 | 50 | 100 | 250 | 500)."
Instead, it should IMMEDIATELY display the disclaimer, and _explicitly_
state that searching is CURRENTLY disabled. How about this?:
"Sorry, but searching is currently disabled.
Wikipedia's internal search capability is disabled during
peak hours (16:00-04:00 UTC / 9:00-17:00 PDT) to reduce
server strain. As a temporary measure, you can search for
text appearing in Wikipedia pages via Google, but be aware that
the index is incomplete and some pages will be out of date.
If you know the exact name of the article, just enter the title's
name and press the 'Go' button at the top right side of the display."
If title searching is being disabled - and not just article title
searching - I think the disclaimer should say that "text search"
is disabled, not "full-text search" (as I've shown above).
Some people (including me!) would presume "full-text search" to mean
searching inside every single article's body, and wouldn't think
that "title search" is the same as "full-text search".
If my understanding is right, this isn't really a "bug" -
it's a user interface issue. I think the current
text could confuse some people (it's certainly confusing me).
By NOT displaying potentially confusing information,
I think it'd be better.
Anyway, I hope I caused no offense. I really appreciate the
work you do!
I don't know if this is related to the other searching problems,
but when the "Go" and "Search" buttons were swapped, some
strange behavior was introduced. "Go" works fine, but
"Search" doesn't. Two problems:
1. An extraneous "search" value in the query is created.
For example, typing "George Mason" and pressing search should try to
GET http://www.wikipedia.org/w/wiki.phtml?search=George+Mason
but instead it tries to GET
http://www.wikipedia.org/w/wiki.phtml?search=George+Mason&search=Search
(note that "search=" appears twice).
Which means you can only search for the word "Search" :-).
2. Even if you hand-fix the search, the title search doesn't work
at all, nor is there any hint that the search wasn't performed.
It should either do a title search (that's not expensive), or
at least say "No title search performed, server too overwhelmed".
I hope that's clear..!
I know this is low on our priorities.
Some time ago, someone suggested a better way of doing interlang links.
The problem is that as more pages are added into a "set", the work
involed to add yet another one increases. By the time you have about 6
languages, there's a high risk of bad linking, and they end up as a
chain rather than a net of links.
I have an intermediate solution which I think would be simple to implement:
: ignore the link made to the *current* language.
ie, on the English wikipedia, ignore an [[en:foo]] link
this would mean the the SAME block of links can be pasted between all
the involved pages. It's not perfect, but it would speed things up for
the time being :-)
See http://www.wikipedia.org/wiki/Czechoslovakia for what I mean.
I've been kicking myself trying to figure out why we're swapping a lot
more under the new kernel than we were before...
Well, it looks like the maximum memory limit is wrong! There's actually a
setting that one has to flip in the kernel config to use more than 1 gig
of RAM... so we're only using half the available memory. Oops! :)
I'll whip up a fresh kernel and reboot.
-- brion vibber (brion @ pobox.com)