23.11.2013 14:12, Mario Weilguni kirjoitti:
> Am 22.11.2013 16:15, schrieb Tom Lane:
>> [ memo to self: never, ever accept another contrib module whose name
>> isn't a plain SQL identifier ]
>
> Well, in that case and since this is a rarely used extension (I guess
> so), maybe it would be the best to simply rename that extension to
> uuidossp (or whatever) and don't make any special treatment for it?
There are a couple of threads about issues with uuid-ossp (AIUI it's
abandonware at this point). If PostgreSQL had a proper PRNG with a
128-bit state it could just implement uuid_generate_v4() function in
core and most people could probably drop uuid-ossp.
I have a branch[1] which implements uuid_generate_v4 in core using
pg_lrand48, but since it only has 48 bits of state it's probably not an
acceptable replacement for uuid-ossp for now.
Is anyone working on a new PRNG for PostgreSQL at the moment?
/ Oskari
[1] https://github.com/saaros/postgres/compare/core-uuid-v4