On Tue, Feb 10, 2009 at 8:09 PM, Angela <beesley(a)gmail.com> wrote:
Yes, it exists already. It has two-way conversion
between wysiwyg and
normal wikitext modes. It doesn't encourage anything that's not
normally done in wikitext (so no odd fonts or random colors), and only
whole numbers are used when resizing images (so no width:199.99729px).
It's not yet a solution for interlanguage links though as it's
mistakenly removing them.
I never realised how badly people over-use html comments, but I tried
it and I finally found a page on which I could load the wygiwyg (guess
I could have just removed them and then switched to wysiwyw mode if
it's all a sandbox wiki anyway but I wasn't thinking of that).
Anyway is there another mailing list or some place to discuss this
tool? If it's all javascript I might be able to help, but otherwise I
would still have a long laundry list of suggestions.
While it would be a fun project for all I agree that it wouldn't be a
good solution to the problem identified by Mark, which is that certain
languages "use a regular colon as a letter", producing too many
collisions with interwiki link prefixes.
English Wikipedia has this problem in a few areas but it is rare
enough that nobody complains about it. For example the search engine
software properly named "ht://Dig" must use the title [[Ht-//Dig]]
(because "ht:" is the prefix for haitian creole language). John
Broughton's international bestseller [[Wikipedia:The Missing Manual]]
is similarly colectimised with a little note explaining why, but it's
just unimportant punctuation to us. It doesn't affect how our words
are pronounced.
It boils down to the inability to create a local page whose title
conflicts with a namespace or interwiki prefix (or to access it if it
existed before the conflict was created [1]), which seems to be a core
limitation of the software. No advancement in wysiwyr syntax-hiding is
going to change this.
I realise there are at least a dozen ways now to visually override the
article title and obscure the fact that it is stored in the database
with something else in place of the colon. That works fine when you're
reading the article itself, but let's take Mark's example again. If
the user types "cu:Next Tuesday", selects it with the cursor, and
clicks the link-me button, is the software supposed to guess whether
they want to link to the (bowdlerised) local title or the (verbatim)
interwiki title?
I suppose it could do a quick check for existence and flip a coin to
break any tie, but I know there are better ways to deal with this.
—C.W.
[1] I recall allegations that one or more articles were stuck in limbo
when "d:" (as in "dictionary") was added as an alias for
"wikt:", but
I don't know the details.