Why would a decent container not be able to optimize that according to the
values in the deployment descriptor (i.e. in the stub code generated
transparently on deployment like e.g. orion does it)? Both cases don't need
to be treated at runtime or am I missing something? If that is correct, I
disagree here, if there aren't other reasons than performance for restricting
local EJBs.
Regards,
Robert
>
> 1.
> The specification should require that Local interfaces use only the
> "Supports" transaction attribute. The Supports Tx attribute simply
> propagates the callers (always another bean for local calls)
> transaction. This eliminates the overhead of intercepting the call for
> the purpose of evaluating the Tx context according to the Tx attribute.
>
> 2.
> The specification should require that the security identity attribute of
> the Local interface be propagated from the caller. It should not be
> allowed to use a runAs security identity. Again this reduces the
> overhead of delegating calls from the LocalEJBObject to the bean
> instance since the security principal doesn't change.
>
> Informed feedback is welcome.
>
> Thanks,
>
> Richard
>
===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
of the message "signoff EJB-INTEREST". For general help, send email to
[EMAIL PROTECTED] and include in the body of the message "help".
