On Thu, Jun 30, 2011 at 11:09 AM, Max OrHai <[email protected]> wrote:

> Thanks for the link! RDP looks quite interesting, and I'm looking forward
> to further developments. Some of the space-time leakage problems of the
> early FRP models have been addressed with Nillson and Hudak's Arrows-based
> Yampa system; could you use any of this in your Haskell RDP implementation?


RDP isolates state to agents and will not have the space leak issues of FRP.
It is, of course, possible to accumulate history in an agent... but it won't
easily happen by accident, which was the problem in FRP.

The Haskell model of RDP is arrows based, using Adam Megacz's 'Generalized
Arrows' to control divergence, delay, and synchronization [1].

But there is no relationship to Yampa's arrowized model. Yampa has a very
heavy focus on 'events', to such an extent it models conjunction with a
stream of sum-type update events. RDP rejects events in favor of duration
coupling of RESTful behaviors, as described in the earlier link.

I thank you for the link to Robin Milner's work. I was not familiar with his
'bigraph' model.

I was initially exposed to spatial modeling through the Kell
calculus. Similar to Milner's bigraphical model, I also pursue distribution
of agents orthogonal to connectivity between them. But, despite deliberating
on this subject many times, I have decided repeatedly to reject an explicit
spatial model. I implicitly get some spatial constraints by modeling
disruption, delay, and synchronization. I can annotate code to direct and
constrain code distribution (e.g. A nearby B). Developers can model space
more explicitly with registries or context objects.

Regards,

David Barbour


[1] http://lambda-the-ultimate.org/node/4259
_______________________________________________
fonc mailing list
[email protected]
http://vpri.org/mailman/listinfo/fonc

Reply via email to