On Thu, 5 Aug 2004, David Garamond wrote:
> Stephan Szabo wrote:
> >>in oracle 10g, you can issue:
> >>
> >> ALTER SESSION SET NLS_COMP = ansi;
> >> ALTER SESSION SET NLS_SORT = binary_ci;
> >>
> >>do you think this is an elegant solution for case insensitive sorting &
> >>searching? is there interest in seeing this in postgres?
> >
> > IMHO, no on both questions. There's always danger on relying on the
> > value of session variables in general in that an application must either
> > set the variable immediately before sending queries that use it (breaking
> > the transparency) or must be willing to deal with the fact that it might
> > not be what you expect. For the second, I don't see how this really does
> > much that the standard spec collation stuff can't do better and I'd think
> > that'd be a much better route to go.
>
> Could you point me where in the archives can I read more? I'm having a
> bit of trouble finding discussion on this. Thanks.
I didn't spend too much time looking, but there are a few that look like
they'll touch upon related issues:
http://archives.postgresql.org/pgsql-hackers/2003-11/msg01299.phphttp://archives.postgresql.org/pgsql-hackers/2001-11/msg00610.phphttp://archives.postgresql.org/pgsql-hackers/2002-11/msg00515.php
And a message where I pulled some text out of the SQL92 draft:
http://archives.postgresql.org/pgsql-general/2003-08/msg00620.php
