On Tue, Oct 12, 2021 at 10:31 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> If the behavior v14 had implemented were "throw an error if the
> first word doesn't match the current database name", perhaps nobody
> would have questioned it. But that's not what we have. It's fairly
> clear that neither you nor Mark thought very much about this case,
> let alone tested it. Given that, I am not very pleased that you
> are retroactively trying to justify breaking it by claiming that
> it was already broken. It's been that way since 7.3 implemented
> schemas, more or less, and nobody's complained about it. Therefore
> I see little argument for changing that behavior. Changing it in
> an already-released branch is especially suspect.
Oh, give me a break. The previous behavior obviously hasn't been
tested either, and is broken on its face. If someone *had* complained
about it, I imagine you would have promptly fixed it and likely
back-patched the fix, probably in under 24 hours from the time of the
report. I find it difficult to take seriously the contention that
anyone is expecting \d dlsgjdsghj.sdhg.l.dsg.jkhsdg.foo.bar to work
like \d foo.bar, or that they would even prefer that behavior over an
error message. You're carefully avoiding addressing that question in
favor of having a discussion of backward compatibility, but a better
term for what we're talking about here would be bug-compatibility.
--
Robert Haas
EDB: http://www.enterprisedb.com