Re: Temp table question

"kyle hailey" <oraperf_at_yahoo.com> wrote in message news:i7p1at8lcsmnhu4aip14lfjreh5nfmb6md_at_4ax.com...
> On Sat, 3 Mar 2001 10:27:48 +1100, "Howard J. Rogers"
> <howardjr_at_www.com> wrote:
>
>
> >And global temporary tables are private to the session that populated
> >them -and a good thing too. One of their main benefits is that run like
> >greased lightening, and they do that by not having any DML locking
> >mechanisms. You can only forego record locks and the like when you can
> >absolutely guarantee that no other user is going to start raising
> >transactions that conflict or compete with your own... and *that* can
 only
> >be guaranteed if the data is strictly private to your session.
> >
> >Regards
> >HJR
>
> Efficient multi access temporary tables could be implemented fairly
> easily in several ways. If the data was kept in the SGA instead of
> process space that would allow efficient and shareable access. There
> would be some latching involved of course but latches are light
> weight. With the data in the SGA

Funny... I always thought the Buffer Cache (and hence, boring old table data that requires careful locking mechanisms to allow concurrent but not conflicting access) *was* in the SGA.

>one could allow other users to see
> the data but not change it.

Funny... sounds suspiciously like a normal data buffer with a row exclusive lock on it.

>Alternatively the first session could
> release a temporary table which could be then obtained by another
> session to be manipulated.
>

Funny... sounds suspiciously like table locks to me.

> Oracle has done much more complicated things in a similiar vein with
> MTS and transaction processing.
>

I think the point is that complicated things are what we have ordinary tables for. Temporary tables weren't designed for that sort of thing, and consequently can forego the complications.

Regards
HJR
> Best
> Kyle Hailey
>
>
>
>
>
Received on Fri Mar 02 2001 - 19:12:56 CST