On Wed, 2009-07-29 at 18:16 +1200, Angus & Anna Salkeld wrote:
> What if you could have objects with the same name,
> 
> BUT you have to "tell" the object what the unique key is.

and how does that differ from special casing everything?

I'll have to allocated a random keyword and add a unique value to it for
each object I add... i also assume this value needs to be the same
cluster wise for sync to work..

> 
> Then you could use a simplified version of an xpath like naming scheme
> (we have been here before).
> 
> /clusternodes/cluster[name=freddy]/bla bla
> /clusternodes/cluster[name=tom]/bla bla

Yeah but this already works with any key.

> 
> I don't like the (currently) completely unstructured approach.
> 
> But as long as we can convert path to handle and handle to path
> reliably and quickly (from within the objdb), then we can have a
> friendlier API (and be more efficient at the same time).
> 
> Would you be OK with that?

I understand where you are going and I am fine to cleanup the API..
I am NOT ok to break xml imports or having to do fancy things to get
objects in there or having to track extra info.
If you need a unique key, then the objdb could do that internally. You
already have object_handles for that...

Fabio


_______________________________________________
Openais mailing list
[email protected]
https://lists.linux-foundation.org/mailman/listinfo/openais

Reply via email to