On Wed, 14 Nov 2007 17:17:48 +0000, Thomas Dalton
wrote:
Parsing
of the pathological cases doesn't seem specifiable, but a
simplified version probably would be.
What if we only allowed ''italic'', '''bold'''
and ''''bold italic'''',
and required a separator between consecutive markup. I.e.
''a''<s/>'''b''' =>
<i>a</i><b>b</b>;
''a'''''b''' =>
<i>a'''</a>b<b>..?
What if we didn't allow nesting, so ''italic and
'''bold''''' had to be
written as ''italic and
''<s/>''''bold''''?
That would probably go along way toward making it specifiable, without
affecting 99% of the current text.
I think it's been agreed that outright rejecting any wikitext is a bad
idea. Error messages or not, the parser has to at least try.
We don't need error messages; just a way to interpret the syntax without
too much lookahead. The combination of ambiguous syntax and nesting is
what makes this hard. It's already been decided that we can't change the
ambiguous syntax. But it seems like things that aren't much used, like
the nesting, may still be on the table.
Why would you remove the nesting, it is highly useful, saves a lot of
time, and forcing it to be done without nesting would confuse
non-technical users. What would be the purpose of removing a useful
feature? The discussions seams to be swaying more towards ease of
documentation\programming rather than usability which should be the
primary goal.
MinuteElectron.