[OpenSER-Devel] discussion: issues with local_route

[Dan Pascu]

On Thursday 12 June 2008, Henning Westerholt wrote:
> So i've came to the same conclusion like Juha. Its better to revert the
> commits that were done so far, upload the patch to the tracker, and
> continue with the developing after the release. People that really need
> this feature probably maintain anyway their own tree, and already
> applying custom patches. I think it makes more sense to err on the side
> of stability and correctness.

Sorry for the late reply, I was on vacation.

I disagree that it should be removed. If it has side effects that are not 
easy to fix during the freeze, then it should be turned into an 
experimental feature that can easily be enabled at compile time by using 
a define.

This new route fixes a very serious issue, by allowing one to enable 
accounting for the internally generated BYEs. This new route may have 
issues, but is solving some issues at hand, not all the issues that could 
be solved by it. That can be extended in the future. I would like to know 
when was the last time when some complete, perfect new feature was 
committed to svn and never changed afterwards? Should I remind you that 
even after about 3 incarnations, the pseudovariable system still has 
issues dealing with NULL values correctly (by design) and that certain 
operators do not work in all cases? What about the TLS implementation 
which is practically useless for end user devices, and even for proxy to 
proxy connections can render your proxy completely unresponsive under 
certain conditions? Or the dialog's module ability to end a dialog that 
cannot be used in practice because one would lose accounting?
Currently there are a lot of things in openser that need tweaking and 
workarounds in the script and none seems concerned about them breaking 
anything or creating confusion.

I would like to see some more constructive approaches about this, with 
people trying to find a way to fix the issues rather that going on a 
chop-it-down frenzy.

I'd say we have a month to identify all the issues, see what we can fix 
and if after that we still have things that do not fit in the bug fixes 
category and cannot be fixed easily but affect the correctness of the 
proxy behavior, then we should turn this into an experimental feature 
that can be enabled by those interested at compile time with a define.
I would also suggest we enable the route gradually for the module 
functions where we know for sure that it works, instead of group enabling 
it on almost everything.


-- 
Dan