database character encoding

What might be best is converting specific columns to utf8 instead of all of them, that way key columns like the file names remain in latin1 and you gain the extra filename length. It looks like MySQL from 4.1 on can set character sets for individual columns, but I didn't find specifics on 4.0 yet.

In that case, I would rather use ascii instead of latin1. Mixing latin1 and utf8 together could be a source of confusion, but using ascii should not cause problems, since it's a subset of utf8.

But notice that having utf8 columns doesn't use more bytes if you only have ascii chars, it only influence size of index fields (like primary key).

Some tables, such as sitellite_filesystem, need all the index they can get. In those cases, it would be best to use latin1 or ascii (I believe those are already hard-coded to latin1 anyway). Otherwise, converting the defaults to utf-8 in the database shouldn't change anything for new sites.

Existing sites would either have to run as-is with no change, or may need a conversion of the data into utf8. I lean towards upgrades keeping their existing character sets so we're not doing anything to potentially mess with data on them.

In this way, new installs would all use utf8 by default (except certain data fields as per the above) and existing sites would carry on unchanged. I think that works best.