Commit experimental features when we branch for 2.20rc1 and tree is unfrozen

[Vlad Dascalu]

Heya,

The two major features that some people have been working on for a while 
are:

-> DB abstractization layer
    -> DB::Compat
    -> PostgreSQL support
-> Email reorganization (the MTA thing)

Since we'll soon branch for 2.20rc1, I suggest to take advantage of the 
3 month period that we're going to have until we feature-freeze for 
2.22. Usually we expect perfection, but on large scale features, this is 
impossible, and those features got stalled for so long, that's it's 
highly unprobable that they will get commited within the near future if 
we continue the established practices in the same style.

I've seen stories before according to which MTA generic support was 
gonna get checked in as soon as 2.18 tree is unfrozen. That never 
happened, despite justdave taking it under his control and stuff.

I suggest to be ready to commit those the same day (or so) when the tree 
opens, so that we can test those on the tip and fix any regression that 
might appear. Taking a defensive approach with those might push them 
after 2.22, well beyond.

By performing an analysis on our support mailing list, a lot of emails 
are related to PostgreSQL support and the MTA thing. I think commiting 
those and letting the community do the testing would be more beneficial 
rather then a draconic review style on the proposed patches with huge 
delays in implementation.

Vlad.