[cap-talk] How desirable / feasible is a persistent OCAP language?
zarutian+cap-talk at gmail.com
Thu Jul 17 16:43:13 CDT 2008
Hmm... interesting discussion.
I think that orthogonal persistence is very handy and co opts the
intuitiveness of 4d space-time.
(What does that mean? Well treat an persistent vat as any other
container in real life.
But instead of putting things made out of matter in it you just put
things made out of information.)
However selective persistence (Java Original Serialization System, E
uncallers, JSON and so on) forces programmers to isolate and abstract
the precious data from the rest of the program/system instead of it
being scattered around the programs process runtime state.
So in my opinion an orthogonal persistence of Actors (similar to
Capros/Eros/KeyKos processes where the Actors address book is just an
map of local names to capabilities) and objects in Vats should be
written with upgrade in mind. (Aka provide serialization methods)
The upgrade method would be something on these lines:
1) Ask the thing being upgraded to serialize/depict itself.
2) Tear down the thing
3) Make an instance of the new version of the thing
4) Ask it to deserialize the depiction from step 1
The new thing could for backward interface compatibility provide
adapters from the
old interface(s) to the new one(s).
On types: to me they are just yet another property of an object.
Just my two zorkmids toward this discussion
More information about the cap-talk