Tom Lane <tgl@sss.pgh.pa.us> writes:
> Greg Stark <gsstark@mit.edu> writes:
> > Is there a reason postgres goes out of its way to pick names that
> > will be harder to work with than necessary?
>
> If we use ordinary identifiers for system-generated names then we will
> be infringing on user name space --- ie, there's a potential for
> conflict. I suppose we could use long randomly-generated names like
> ewjncm343cnlen, but are those really easier to work with?
I don't see an unseverable link between "user name space" and "identifiers
that don't need to be quoted". Mixed case names for instance seem like
perfectly good user name space identifiers.
Postgres could just as easily say "the system reserves all identifiers
starting with $" and still not require quoting $.
> I think a more useful approach is to treat it as a documentation
> problem. Perhaps an example in the ALTER TABLE man page would help.
--
greg