Re: 'TID index'

On Sat, Sep 25, 2004 at 11:14:53AM +0100, Simon Riggs wrote:
> >Jim C. Nasby
>
> Fair questions. MVCC has been tightly locked into Postgres/SQL for the whole
> of its history. There is much written on this and you should search some
> more - references are in the manual.

Well, not quite it's whole history: MVCC showed up in 6.5. Vacuum's been
there since before SQL. Actually, is a bit of a historical accident. My
understanding of the squence of events is that Hewlett-Packard donated
an early WORM optical drive to the Stonebraker lab. Since it's write
once, it had the beavior that you could only append to files. Someone
thought it might be useful for auditing, etc., so they wrote the first
storage mananger for postgres to accommodate that drive. The other
storage manager at the time was for battery-backed, persistent RAM.
So, all this append-only writing leads to files with lots of dead
tuples, so the vacuum command was added to reclaim space. I think on the
WORM drive, this was supposed to mark blocks 'invisible' in some sense.

I don't know if the WORM drive ever actually got used with postgres.

Ross

--
Ross Reedstrom, Ph.D. reedstrm(at)rice(dot)edu
Research Scientist phone: 713-348-6166
The Connexions Project http://cnx.rice.edu fax: 713-348-3665
Rice University MS-375, Houston, TX 77005
GPG Key fingerprint = F023 82C8 9B0E 2CC6 0D8E F888 D3AE 810E 88F0 BEDE