Re: cataloguing NOT NULL constraints

On Wed, May 8, 2024 at 4:42 PM Alvaro Herrera <alvherre(at)alvh(dot)no-ip(dot)org> wrote:
> I spent a long time trying to think how to fix this, and I had despaired
> wanting to write that I would need to revert the whole NOT NULL business
> for pg17 -- but that was until I realized that we don't actually need
> this NOT NULL NO INHERIT business except during pg_upgrade, and that
> simplifies things enough to give me confidence that the whole feature
> can be kept.

Yeah, I have to admit that the ongoing bug fixing here has started to
make me a bit nervous, but I also can't totally follow everything
that's under discussion, so I don't want to rush to judgement. I feel
like we might need some documentation or a README or something that
explains the takeaway from the recent commits dealing with no-inherit
constraints. None of those commits updated the documentation, which
may be fine, but neither the resulting behavior nor the reasoning
behind it is obvious. It's not enough for it to be correct -- it has
to be understandable enough to the hive mind that we can maintain it
going forward.

--
Robert Haas
EDB: http://www.enterprisedb.com