> You seem to have fairly strong views on this subject.
On that subject? Yes. The irony, Aaron, is that *I* was the one who
suggested, last year, that JNDI might be our universal access method.
Harmeet picked up that idea from me. So when I say that it is a
spectacularly bad idea, remember that the primary criticism is aimed at me.
I believe that JNDI can perform as we need for the user repository. We
would likely have to write a JDBC Service Provider for JNDI, since I haven't
found one yet (IBM has one, but they don't want to share :-)). And I still
want to put Hypersonic SQL into the standard build so that we can count on a
SQL engine for high duty cycle lookups.
In any event, the user repository will be one area to deal with.
Serge suggested JavaMail as the message stores interface. I'm not sure what
the performance is, but we do have the JavaMaildir author to help out, and
I'm sure that he is familiar with the performance.
Also, please note that we're not considering JavaMail for the spool. The
spooler needs to be high performance, but it will only move the Mail object,
which have a looser reference to the MimeMessage. The MimeMessage wouldn't
move as much as it does in the current scheme, cutting down on that
overhead.
> Until we have found the silver bullet, it might be
> better not to throw anything out?
We can eliminate what won't work. JNDI as a message store interface won't
work. JavaMail is the current best candidate.
All of the above represent my views, of course. Hopefully Danny, Serge, et
al., will opinion soon, along with others, so that we can reach a consensus,
and get people coding.
--- Noel
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
