[Mediawiki-l] iso-8859-1 conversion to UTF-8 failed during upgrade from 1.4.0 to 1.5.2

[Brion Vibber]

Andre Oliveira da Costa wrote:
>> I had the same problem, when using the upgrade1_5.php and update.php
>> scripts. After I put this into my LocalSettings.php it worked though: 
>>
>>   $wgLegacyEncoding = 'ISO-8859-1';
> 
> Mmmh... this looks promising. But with this setting did you need to run
> upgrade1_5.php after all? Or did this simply tell mediawiki to "accept"
> iso-8859-1, and you left your DB untouched? 

$wgLegacyEncoding tells the wiki to apply upconversion at load time on 
old text entries that don't have the 'utf-8' flag on them.

It does _not_ apply any conversion at run time for page titles, 
usernames, edit comments, etc. These are what upgrade1_5.php will apply 
transformations to if you have the reencoding trigger on. (This was 
created primarily for internal use at Wikimedia and is not well 
documented, sorry. Read the source code to check for configuration options.)

> If I were sure this setting is going to be around forever, I would
> probably go this way, too. But, I am affraid iso-8859-1 might not be
> supported at all in a near future and, in this case we would just be
> putting off an inevitable upgrade to UTF-8... Do you know if this
> setting is official or is it an "undocumented feature"?

 From 1.5 on *only* UTF-8 is supported.

$wgLegacyEncoding for upconverting old text entries (not anything else, 
just text entries) will continue to be there for the forseeable future.

-- brion vibber (brion @ pobox.com)