[dev] RDO describe() limitation

[Chuck Hagenbuch]

Quoting Duck <duck at obala.net>:

> Oky, to not make confusions I will change it to Minerva, but I guess  
> the usage is a common case. We have an central “index” table with  
> main invoice data,
> and other related data tables for articles, taxes, currencies etc.
>
> If invoice is edited, is done like in datatree.  The related data are
> completely deleted and then reinserted. Even only one article price or
> quantity is changed. Whit a change like this the RDO will update only
> actually only the changed article. Performing the update much faster, with a
> much less code, and more safety if something goes wrong with the dataserver.

Okay, so having read:
http://framework.zend.com/wiki/spaces/viewthread.action?key=ZFMLGEN&id=2080

and then coming back to your email, I'm actually not sure if you're  
arguing for always using surrogate keys (thus making sure you can  
always refer to an individual row) or for not requiring surrogate keys  
- i.e., allowing a key to be multiple fields.

I think you've convinced me, in combination with that thread and the  
postgres thread linked from it (very long) that we should support  
multiple-column keys. But I haven't looked at implementing it yet.

-chuck

-- 
"we are plastered to the windshield of the bus that is time." - Chris