Otherwise backends which do not rely on sql can't use many-to-many relations as discussed here:
​http://groups.google.com/group/django-developers/browse_thread/thread/c3cd2eb4f63aed5d/6e8de9e05424d18f
(do not care with messages from people that doesn't know how to use reply-to on mailing-lists)

Now let's discuss around the proposed patch.

Initial patch, without any documentation nor tests, design discussion for now

This needs to happen as part of a bigger refactor of related fields.

Noting here that this(or some variant thereof) is going to be a blocker on having many to many fields work correctly with Multiple Database support(GSOC).

Update patch against r10618

For the record:

[15:43] <Alex_Gaynor> david`bgk: bitchin! 
There's probably no time for that to go in before 1.1. 
But if you remind me I'll bug the shit out of jacobkm 
and russ to get it in fterwords

I'll ping you anyway.

See #11795 for a related request that this ticket may make possible.

RC1 of the m2m refactor stemming from Alex's GSoC 2009 project

The patch I have just uploaded is ready for trunk. The code is also available on ​Alex Gaynor's github branch

My intention is to commit this mid next week. Feedback is welcome.

RC2 of m2m-refactor patch. Incorporates fix for abstract base classes.

Another small issue: Intermediary models should have 'unique_together': (from_, to), to match ​the old behaviour.

RC3 for m2m-refactor patch. Incorporates unique_together fix.

(In [11710]) Fixed #10109 -- Removed the use of raw SQL in many-to-many fields by introducing an autogenerated through model.

This is the first part of Alex Gaynor's GSoC project to add Multi-db support to Django.

Milestone 1.2 deleted