Brion Vibber wrote:
The second pass reads this stub dump, and the previous full dump of the same database. Existing revision text can be copied directly from the previous dump (page contents on a given revision ID are immutable).
The thing I completely forgot to mention about this is that I'm using the new XMLReader extension in PHP 5.1 for this; so srv35 and srv36 have experimental PHP 5.1.0RC1 installations in /usr/local/php5 that get used for this step.
XMLReader has a 'pull' interface, so you can read off the XML stream at your own pace. Quite handy when you're already trapped in one SAX event loop reading the first stream. :)
-- brion vibber (brion @ pobox.com)