"Ann W. Harrison" <aharrison@...> wrote:

>
> But is the feature either powerful or fast? I think not.

And here we disagree.

> What people coming from Sybase/MSSQL want are not what the
> standard calls "temporary tables." We can do that, but
> we shouldn't pretend that using the language of the standard
> makes a non-standard feature standard-compliant.

First of all, we've already almost agreed on GTTs. This is what the most
vendors offer, this is quite powerful feature (I used GTTs in Oracle) and
proper design/implementation can make it fast. Let's just solve the
remaining issues and do the job.

Secondly, I don't actually care much about those from Sybase ASE or MSSQL,
but they also implement GTTs which just extend our (and standard) semantics
by allowing GTTs to share their data between sessions. Those who rely on
this extention are going to be in troubles with Firebird, but others will
benefit from our implementation as well.

Then, it seems that only Sybase ASE / MSSQL implement LTTs. Regardless of
what the SQL spec says this is exacty what people expect from LTTs -
session-level visibility. I don't know whether such a semantics is valuable
or not, but at least Volker thinks so. So it's worth discussing more.
Perhaps, you Ann could ask the core PostgreSQL or MySQL guys about their
position on this issue. I'd very much like to know what exactly stops Oracle
/ Sybase ASA and PostgreSQL from offering LTTs - misunderstanding of the
standard or any technical issues or just lack of value.

As a personal note, I'd stick with <created LTTs> as session-scope objects
and <declared LTTs> as routine-scope objects to make everyone happy. This is
one of those very rare cases when I avoid following the standard exacty and
treat it the way it's valuable for us and our users.


Dmitry