axis-java-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Glen Daniels" <>
Subject Re: Session support
Date Sun, 27 May 2001 20:46:09 GMT
> This is more mechanism than I will build in my first submission.  The net
> effect will be the same, but I plan to have *only* a simple ID in the
> message context, and the ID-to-target-object table will live in the
> RPCProvider (which I believe is more consistent with how, for example,
> work).  Glen, if you want a more bodacious SessionContext, you are free to
> write it :-)

Natch, and I likely shall (though Doug is free to -1 it :)).

The thing I'm concerned about is that with just the ID in there, every
Handler who wants to deal with session-related information will have to have
their own mechanism for relating that ID to some backend data store.  I
don't think that Providers are going to be the only Handlers that will end
up being session aware.

> >With a system like this in place, you could also write a simple
> >session management system which uses SOAP headers to carry session-ids,
> >which tie to Hashtables that are pooled and expire.  Any
> >Handler/extension/backend who wants to store stuff in the session bag
> >doesn't need to care how the actual session is being managed.  The
> >RPCDispatch sample code above wouldn't change a bit.
> This will be just as true in the code I next submit.

Hm.  OK, that seems a bit at odds with your last paragraph.  I mean, I can
see how with just an ID you could write a Handler which maps that ID to
(say) a target-object field in the MessageContext, but it didn't sound like
you're planning on any more general mechanism for accessing session-related
data (i.e. "store stuff in the session bag").

> My submission will support Apache SOAP-like SESSION / APPLICATION /
> lifecycles for Java providers, and the next submission will hook in
> for HTTP servlet session mechanisms.

I'm psyched to see it!  Did you get your CVS access working yet?  When do
you think this might be ready - it's got the ServiceClient stuff in it too,
right?  I've been holding off on doing a few things relating to type
mappings and dispatch until that abstraction goes in.


View raw message