On Mon, 2007-05-14 at 09:33 +0200, Peter Rabbitson wrote:
> Paul J Stevens wrote:
> > Aaron Stone wrote:
> >> We issue only a "SET NAMES 'encoding'" without the COLLATION argument,
> >> so at the moment we fall onto whatever default the database has, but...
> >>
> >> Unless Paul changed any queries with the major utf8 cleanup today, the
> >> database collation doesn't matter because we don't ask the database to
> >> do any sorting on utf8 columns; all of the sorting takes place on the
> >> DBMail side in concert with GMime.
> > 
> > No queries were changed. The cleanup wasn't that major actually. But the
> > collation does matter because in the IMAP SORT code we do rely on the 
> > database
> > for collation and sorting. It seems to me that's a good thing actually, 
> > because
> > it allows us to let the database handle this issue.
> > 
> > We should probably change the table definitions to use UTF8 encoding and
> > utf8_unicode_ci. DBAs can ofcourse change collation per table or per column 
> > as
> > desired. Dbmail doesn't really care at the moment.
> 
> Erm... It does. It detects collation mismatches just as encoding
> mismatches. See the thread titled 'mysql collation for dbmail'. Or it
> might be that I am using an antiquated dbmail version? (the latest
> 2.2.5-0rc1 debian package)

It's equally possible that I was being too pedantic about looking for
anything resembling an encoding, and the best fix is to remove the
collation check. I have not looked deeply into these issues recently, so
I'm scratching to find out if someone really knows the right thing to do
in this situation.

Aaron

_______________________________________________
DBmail mailing list
[email protected]
https://mailman.fastxs.nl/mailman/listinfo/dbmail

Reply via email to