Thought I'd post here since I think everyone drops by this forum...

Dan Wickstrom suggested that limiting the maximum numbers of immediate
descendents of a node in our trees to 32768 would be too limiting,
particular for the acs_objects table which has a flatter topology than
I'd imagined.  He convinced me that he's right.

So I've changed the PG datamodel to use 32 bit keys (we still use one
byte for short keys).  The actual ranges are 0-127 in a short key, 128
-  2^31-1 in a long key.  That should be plenty of immediate
descendents - somewhat over two billion of them allowed per node.

This only required some code changes in postgresql.sql but does
require that you reinstall.

As mentioned in the subject line I do believe this to be the last
widescale change to the core datamodel for this release cycle.

Sorry for any inconvenience but we want to get this stuff *right*
before we issue a beta and then final release.

Right.  This is PG only.  Oracle's been quite stable and will remain so...

Don, while you're giving an update on major changes, could you
also please tell us which modules are still likely to undergo
major bug fixes or modifications before first release and which
ones you are confident are mostly ready?

Jun - you'll have to write custom export/import sripts I'm afraid.

If the sites you have are working reasonably well, and you don't expect a huge growth in users, you might just leave them alone.  On the other hand, if you think you'll ever have to upgrade it, you're probably better off doing it ASAP.

Michael - for the most part I think packages are pretty stable.  I don't expect much change at all (I think getting Neophytos's improvements to search are important, though).  Folks at Jun's company have been working on doing some additional porting but I don't know if they'll get done in time for this release (I need to check with them soon, I guess).