Ok so my current toughts are:

* Use one MailboxMapper/MessageMapper/SubscriptionMapper per request.
Which in fact should share the same JCR Session / EntityManager etc. a
request should be handled by a "new" thread so we would have a
ThreaLocal like pattern. The instance should get stored in the
MailboxSession while processing the request.

* Dispose the MailboxMapper/MessageMapper after each request to be
sure that the underlying Entitymanager/JCR Session get closed. If we
would close it before we could get into trouble when lazy load stuff
etc.

Does this sound good ?

Bye,
Norman


2010/5/5 Robert Burrell Donkin <[email protected]>:
> On Wed, May 5, 2010 at 12:52 PM, Norman Maurer
> <[email protected]> wrote:
>
> <snip>
>
>>>> Would it be worth discussing where James IMAP is going in the near-term
>>>> and whether we should coordinate.
>>>>
>>>
>>> You are totally right. I'd be glad if we could find a good design for IMAP
>>> and implement it as soon as possible, not least because my maildir
>>> implementation will be affected by this process.
>
> IMAP is a subtle protocol. designs are non-trivial.
>
>>> I'd like to build on a relatively stable structure.
>
> use an internal API, then any changes can be bridged. a good
> commons-style Maildir library would be really useful generally and
> much easier to debug. so that's where i'd start.
>
>>> That would, by the way, be my only objection to
>>> releasing James too soon. First we should stabilize the IMAP stuff.
>
> release often, release early ;-)
>
> better to ship something with a low version number
>
>> Me too. I did a lot of refactoring over the last weeks and sometimes I
>> feel like just turning cycles..
>
> +1
>
> always seemed like that to me as well...
>
>> I the long term I would like to see
>> some kind of async processing for mailboxes. But I think thats not
>> feasible in the short term.
>
> dunno
>
> the reworking i did always had async in mind. last time i looked,
> should be reasonably easy to flip into the message passing paradigm...
>
> - robert
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [email protected]
> For additional commands, e-mail: [email protected]
>
>

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to