[OpenGuides-Dev] Wikiness - discussion on london.pm mailing list - OpenGuides-Dev

23 Jul 2004

I'd welcome feedback from OpenGuides people as well.

...
  -----Original Message-----
 From: london.pm-admin(a)london.pm.org
 [mailto:london.pm-admin@london.pm.org]On Behalf Of Tim Sweetman
 Sent: 23 July 2004 15:02
 To: london.pm(a)london.pm.org
 Subject: Re: wiki-ness (Was: Re: devolving and crack fuel)

 .....

  if you choose the right 
 software you can make them go away.
 http://search.cpan.org/~kake/CGI-Wiki-Kwiki-0.52/ is quite   nice when you 
  make it use usemod formatting and make it not use
studly   caps for links.

 I think there's a basic conflict here:
 1. authors want to write links like
 [shepherds bush]
 ShepherdsBush
 [shepherd's bush] and have them work.

 StudlyCaps are just plain easier to type, and avoid doing
 [this

 or

 ]this, and confusing the parser, and the writer, who's not that 
 technical and isn't used to incorrect syntax blowing the 
 bloody doors off.

 2. readers want stuff to be formatted like English (if they 
 care, which 
 they might not) and to avoid miscontextualisation along the lines of

     ... the bodyguard shepherds Bush through the throng ...
     (not saying if it was Kate, Vannevar or a George, mind)

 Punctuation helps. StudlyCapsLookAwful, especially to 
 nonprogrammers who 
 are just plain not used to them.

 My solution would be:
 a) pages can have multiple names, one of which is the 
 preferred/canonical one. (This is a bit like the more widespread 
 "redirect node" feature)
 b) links are written with the node's canonical name.

 writer writes ShepherdsBush (or [shepherd's bush])
 reader sees Shepherd's Bush.

 (This could work well with people/entities who insist on 
 spelling their 
 names in a "difficult" way, such as with accents your 
 keyboard doesn't have)

 3. Names may themselves want canonicalisation. Is it Earl's Court or 
 Earls Court? Probably depends which sign you read, and with 
 place names 
 even LynneTruss is bound to get it wrong sometimes.

 No, I haven't implemented this. 
I raised this on the OpenGuides-dev list last year, see
http://openguides.org/mail/openguides-dev/2003-October/thread.html#91

I'd be very interested if it is possible to resolve these issues - at least
arriving at a sensible compromise. OpenGuides and its intrinsic search
function could do with having this matter resolved.

This is how it works at the moment:

1. Node names are normalised before being inserted into the database. This
involves turning spaces to underscores, capitalising initial letters (Title Case), 
and escaping out of pathological metacharacters that would otherwise break the
parsing.

2. Links are normalised as they are followed, permitting use of unnormalised text
e.g. different case.

3. Punctuation characters in titles are preserved (e.g. apostrophes, hyphens)

4. Permitted search strings have a restricted character set

As keeper of the search, I would like to know if we have got this right or not,
especially the normalisation aspect of this.

We are also being faced with new challenges, regarding character sets and
internationalization.

Your feedback would be most welcome :).

Ivor.