On Fri, 10 Aug 2007, Jonah H. Harris wrote:
> The main difference is that PostgreSQL is with-REDO/no-UNDO because it
> stores every row version in the main table, and Oracle/InnoDB are
> with-REDO/with-UNDO and they reconstruct a block and/or row image from
> the log to provide read consistency.
I'm finding it hard to parse this sentence; it's a bit too run-on and I'm
not sure exactly what you mean here. Any chance I could get you (or
someone else who actually understands the distinction you're making better
than I do) to re-write it so it's clearer?
Regardless, I just incorporated all of Jonah's comments from that very
helpful message into the page, and there's a pile of additional references
in there on this topic now. I'm comfortable that Lukas's warning not to
oversell the PostgreSQL MVCC has been addressed.
As part of the final cleanup I'm doing today (in prep for the conversion
to techdocs) I also added a "Credits" section that specicially recognizes
all the work Chris and Lukas did by writing parts of the document. I
assume you two would want your names and links to your web sites in there;
if you want to avoid collateral damage from the seeming inevitable flames
publicizing this info will produce you can certainly make yourself
anonymous instead.
--
* Greg Smith gsmith@gregsmith.com http://www.gregsmith.com Baltimore, MD