04:20 pm, 13 Jan 05
two more wikipedia remarks
I was thinking about some toy projects for Wikipedia data again and I poked around some Wikipedia code. Two observations:
- Good'ol timwi strikes again:
Revision 1.160 - (view) (download) (annotate) - [select for diffs]
What a great develoment model. :(
Sun Aug 1 01:57:26 2004 UTC (5 months, 1 week ago) by timwi
Branch: MAIN
Changes since 1.159: +4 -7 lines
Diff to previous 1.159
Since nobody commented on this, I assume all the changes are OK
(http://mail.wikipedia.org/pipermail/wikitech-l/2004-July/011789.html)
Use wfMsg's own replacement functionality instead of str_replace ( "$1", ... ) - From a recent change:
BUG#1317 - External links in image caption not working
I know it's against the spirit of wikipedia to actually have some careful design go into it, but they would've saved so much pain had they designed wikitext as a regular language instead of a pile of regular expressions. Since they allow HTML in entry bodies, the only way to properly display wiki data is with an HTML browser. :(
Changed regexp to allow single ] inside of an image's alternative text
If they're going to allow HTML anyway, they might as well just use meaningful HTML markup. Why mix together two models? I guess I'm really just saying that because I hate wikitext, though.
That is true, but I'd probably want a more rich language than that if it was to be a source language for transforming into various presentations.
The HTML on Wikipedia isn't that complex, is it? It's not really hard to translate simple HTML to LaTeX. It just gets tricky when people start abusing it for various formatting/display tricks.
In principle I like the idea of simple markup languages like wikis use. Not many of them do a decent job of it, though (what's with the '''multiple quote''' stuff the original wiki uses?).
I use Textile on my blog, and it's really nice, except for when it just mysteriously doesn't do the right thing ... because it's really just a pile of regular expressions instead of a real syntax, so it has bugs and ambiguities galore.
We'll be able to do everything WYSIWYG soon, right?
WYSIWYG, while more convenient for end-users, means we're more likely to lose out on meaningful markup like titles and subheadings. (Unless our WYSIWYG doesn't allow changing font properties...) That data is useful both for creating on-page "table of contents" bits as well as search engines.
That's actually a really good UI idea. A very simple interface, which only allows you to change body levels, and or the formatting for each... it sounds Apple-clean.
And, looking it up Pages looks like that. Go figure.
I came up with a different acronym years back for WYSIWYG but with content-related options rather than presentation-related options. I can't remember what it stood for now or how exactly it was spelled, but it was pronounced “wimmy-weeoo”. I'll have to ask some old friends if they can remember what I was rambling about at the time…
The best implementation of this I've seen so far is the XMLMind XML Editor. Although its UI is a little wacky and could do with a little attention, it basically just takes a DTD and a CSS stylesheet and helps you write a document which satisfies that DTD. I've used this in the past to edit DocBook articles before and it worked pretty well.
The wikipedia site run Perlbal for a bit today ... they're trying it out, because they're sick of Squid and Pen and stuff. Not sure I fully understand, but I think it's awesome they now use memcached and Perlbal, and are considering MogileFS for their growing image collection.
BOO YAH.
vBulletin is a piece of forum software, written in PHP. It's run at somethingawful.com. Recently, the load average on the servers was stupidly high, and they just found the cause of it.
vBulletin uses emoticon codes like :hello: :google: (Many user defined ones). Somethingawful has around 237 smileys.
How they are replaced for the final output is for each smiley, running a search and replace on the posts. They would get processed over 200 times before displaying each post on a page.
Heh.
vBulletin uses emoticon codes like :hello: :google: (Many user defined ones). Somethingawful has around 237 smileys.
How they are replaced for the final output is for each smiley, running a search and replace on the posts. They would get processed over 200 times before displaying each post on a page.
v2.2.9 is a very old version of vBulletin and 237 smilies is an excessive amount. Yes, in 2.2.9 smilies would be replaced on every page. Using v3, there is an option to cache posts (for a defined timespan), which does the replacement for smilies only once.
Out of curiosity - where is the cache stored?
And does version three use a one pass algorithm instead?
Don't know the answers to your other two questions - I asked over in their forum after seeing your comment as that'd definitely affect my decision to purchase. I'm sure they'll be happy to answer pretty quick if you ask them [link]? The developer responded to me in a few minutes (that subforum allows registering by anyone) so they seem on top of things.