Re: Fwd: 2 questions: is alice alive? oz ffi?

Lucas Dixon schrieb:
> I'm hoping that some of the features of Alice, and perhaps some of
> Andreas' work on module systems also, will make it to PolyML.
> 
> hopefully!
> 
> What are the most important features of Alice for you?

Alice did not implement the ideal I'm seeking, but it came a long way. I 
have been waiting for the system to mature a bit more before using it in 
anger, but it never progressed (now I know why), so I have to admit that 
I never used it.

However, I found several ideas very intriguing and have been hoping to 
come around to using them.

I found futures an interesting concept. I have no idea how well they 
work in practice, but they were looking well thought-out.

The other point was a whole area, the way Alice was doing networking. 
The ability to marshal anything at a whim would be very valuable - one 
could do all sorts of load balancing, both automatic (load off some 
object network to a less overworked server), by design (experiment with 
different load distributions, simply by creating the root objects of the 
subnetworks in different servers), and manually (let the user move parts 
of a system around).
What made Alice's approach far more limited than necessary was that it 
wouldn't pickle (marshall) "resources", i.e. things like file 
descriptors. Also, IIRC Alice would simply clone mutable values when 
sent over the network - this is a change in semantics, there are now two 
equal but not identical objects where there was only one before. The 
solution is the same for both cases: don't copy the object, create a 
proxy on the target system that will route all calls back to the source 
server. (You'd want functions to "relocate" some kinds of resources, 
i.e. copy their values to the target and install the proxy in the 
source. Preferrably for entire object networks at once.)

Hope that's food for thought :)

Regards,
Jo