2009/12/1 Marcus Buck wiki@marcusbuck.org:
Tim Starling hett schreven:
Gerard Meijssen wrote:
Hoi, Given that we should be moving forward not backward, it makes more sense to provide Unicode 5.1 characters and webfonts.
The big thing of MediaWiki was that it supported Unicode when this was still a new thing to do. We should support the latest and the best Unicode support.
You did read the post didn't you? Forcing everyone to buy Windows 7 is not generally the way we do things. Unless the client situation is not as bad as it sounds, we will need to have a transition period where we support older clients until their market share falls far lower than 50%, which is where, by Praveen's figures, it is now.
I guess you are both right. To me the best solution seems to be: accept both as input (obviously), normalize everything to 5.1 and store it in that codeset (so our data is consistently 5.1). For output convert it to 5.0 to evade problems with clients not yet ready for 5.1. The advantage is, that our data is stored in the most modern format, but still the clients are served data that they can process. If there are performance problems with the conversion on serving or anything like that, of course storing the data in 5.0 is still good enough. More important than discussing the specific technical details is actually doing it, implementing it.
This problem seems closely allied to Unicode normalization of Hebrew and Arabic where we chose to go with the official standard thus breaking at least most current Microsoft installations at the time which had fonts designed for a different sequence of letters and modifiers.
The code would logically belong in the same place I suspect.
See:
Unicode normalization "sorts" Hebrew/Arabic/Myanmar vowels wrongly https://bugzilla.wikimedia.org/show_bug.cgi?id=2399
http://www.mediawiki.org/wiki/Unicode_normalization_considerations
Andrew Dunbar (hippietrail)
Marcus Buck User:Slomox
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l