* Tom Lane (tgl@sss.pgh.pa.us) wrote:
> It's essential IMHO that we provide pg_shadow and pg_group as reasonably
> backward-compatible views on the new pg_roles catalog. It's not at all
> negotiable that CREATE USER and CREATE GROUP have to still work in a
> sane fashion --- to say otherwise is to say that we aren't going to load
> pg_dumpall scripts from older PG versions, and that is a Non Starter.
Right, makes sense, I had just busted them while getting the initial
code written. I've now gone back and cleaned up the main parser quite a
bit with regard to create/alter/drop/etc user/role. My latest work is
available here:
http://kenobi.snowman.net/~sfrost/pg_role/latest-role_20050609.1.patch.gz
Also the .h files in the same directory (pg_auth_members.h, pg_authid.h)
which need to be put into src/include/catalog/.
It patches and compiles cleanly against latest CVS (at least, latest as
of a few hours ago). I've also updated and flushed out a bit the set of
milestones/todo items. My latest version of that can be found here:
http://kenobi.snowman.net/~sfrost/pg_role/role_milestones
* Means completed in the patch, ? means I'm not sure if it's something
that should be done or not. No marker means it needs to be done and
hasn't been yet. In general I feel it's starting to get close to meeting
all the expectations that I had for it. The more critical things, imv,
are the ACL changes for multi-level role resolution (for owners and
others) and the per-backend role-member cacheing and fixing the other
parsers (ecpg, etc, shouldn't be too hard now that I've got it figured
out for the main parser, or at least think I do).
Unfortunately, it's starting to get close to July 1st and my availablity
is rather sporadic in terms of when I can spend time on this. I'd
certainly be willing to work with others (I'm generally pretty
responsive to email) to get this finished off and polished up. I do
hope to spend some more time on it over the next two weeks and may be
able to finish it by July 1st myself but I can't really be 100% sure on
that.
Thanks,
Stephen
