[Cosmo-dev] sharing update 12/19

[Phillip J. Eby]

At 06:56 PM 12/19/2006 -0800, Brian Moseley wrote:
>from the last thread we had on this exact same subject, it was my
>understanding that we are defining a single valued primary key for
>every record; that it would NOT be a field but rather a property of
>the record itself;

It's a property of the *recordset*, not the record, and it was added to 
support both Cosmo and doing efficient diffs on the Chandler side.


>and that it would be called uuid, since at least
>for sharing, for preview, we are only dealing with transmitting items.

Yes, it's called uuid, and it's a non-namespaced attribute on the 
recordset.  The uuid attributes on individual records are *not related*.


>if i renamed eim:uuid to eim:pk and otherwise left it exactly as is,
>would you have any remaining objections for preview?

You misunderstand.  I'm just pointing out that uuid isn't a namespaced 
attribute on any element.  It's a known attribute of the recordset and 
applies to all records contained in the recordset.

Please see what Morgen implemented, as it is exactly what we discussed 
before, except that he switched to using attributes for simple field values 
(which I understood was only after consulting with you).


>>They include some EIMML samples that should make it easier to see where the
>>spec is out of sync w/Morgen and my vision of how EIMML works.  (They also
>>demonstrate the generic sync/diff algorithms, although not in terms of XML
>>effects.)
>
>actually, morgen and i have discussed in irc over the last several
>days everything that went into this spec, so the xml in that EIMML.txt
>looks quite a bit dated.

Sounds like someone hasn't been keeping me in the loop, then.  :(