28 Nov
2008
28 Nov
'08
10:32 a.m.
Daniel-Constantin Mierla schrieb:
Hello,
there are couple of modules that change the R-URI and automatically call
append_branch() if the execution takes place in failure_route, other
modules don't.
Perhaps we should get to a coherent behavior, if not for existing code,
then for the future. Probably the most flexible, better in my opinion,
is to let script writer to call append_branch() in config -- once done,
it is no way now to revert. A new branch should be automatically done
only if it is a specific operation, e.g., location lookup with more than
one record.
What do you think?
IMO the append_branch() is PITA since ever (ok, it is better since
branch routes) - so a cleanup is a good idea.
From flexibility point of view having the append_branch manually is a
good idea - although it of course will confuse beginners.
IIRC there were also differences between manual and automatic (e.g.
dispatcher) regarding other branch parameters which were not copied
(duri, send_socket, bflags ...) - this should be reviewed too.
klaus
Cheers,
Daniel
PS. this blending of the ser-kamailio sources gives me ideas to do some
cleanup as I review the code ... especially after midnight ...