The MySQL 4.0-backwards-compatible default schema causes a lot of problems on MySQL 4.1 and higher:
- confusing results when doing direct queries
- difficulty sharing data with other applications
- data corruption with 'mysqldump'
The 'MySQL 5'-mode schemas avoid these problems, and are finally working properly with the standard updaters:
- The binary schema preserves full compatibility with non-BMP Unicode characters (and thus data from Wikimedia sites)
- The UTF-8 schema is limited by MySQL's BMP-only support, but still seems reasonably functional.
However, the old schema is still the default, and there's currently no provision for cross-upgrading from the old schema to one of the new ones. The installer/updater should be able to detect the old-style schema and convert. A converter maintenance script might also be helpful.
At the worst, going row-by-row and doing UTF-8 and double-UTF-8 checks and converting line-by-line should do the job. :)
Once conversion is in place, we can consider using the more capable binary schema as the default.
Version: unspecified
Severity: enhancement