Paul J Stevens wrote:
Matthew O'Connor wrote:
I think that is a rather drastic statement and not what I understood.
There are problems when a message has a header that is UTF8 encoded, but
I believe that is against the spec isn't it?
Yep. Headervalues are /supposed/ to be us-ascii, but 8bit strings are
increasingly common (anyone using outlook?), so we will have to
accomodate them. Also, supporting utf8 is just the-right-thing to do.
Late in the 2.1 sequence I experimented with casting the encoding to
utf8 for all caching table values. This led me to conclude that such an
approach is flawed.
By storing values as-is in encoding neutral fields will allow us to
defer conversion to when we actually need it. We can do that relatively
easily and without triggering any regressions.
Does that mean converting all the header caching tables to bytea rather
than text also? I think that makes sense.
_______________________________________________
DBmail mailing list
[email protected]
https://mailman.fastxs.nl/mailman/listinfo/dbmail
