Semantic-based linking

[Francois-Rene Rideau]

Semantics-based linking reminds me of "semi-static binding",
as explained in some paper on the Scheme repository (see Languages Review).
	http://www.cs.indiana.edu/scheme-repository/

It also reminds me of the little I've read about MS' Intentional Programming.
	http://www.research.microsoft.com/research/ip/

Or about problems of delimiting objects during Migration.
	http://www.tunes.org/Migration/

> The principle in tunes is to have standards, but they are formed by
> agreement of all parties (communities of users), not by us (system
> designers).
	We'll propose (so will other designers), the public will dispose.

> Global unique IDs.
Globally Unique IDs, not global IDs.
Globally Unique means that an ID is used by only one object in
the whole world.
Global Unique means that we have somehow(?) unique IDs, with a "flat"
address space over the whole world, so that every one uses the same ID scheme.
This might be a bad thing, since we can optimize lots of things with
Local IDs while preserving the "Globally Unique" property:
if we take the context into account, we can use shorter nicks when possible
(i.e. most often) and do path-prefixing dynamically (with reductions when
possible) when a name enters/exits a given "context" (Luca Cardelli calls
them "ambients").
	http://www.luca.demon.co.uk

Also, unicity doesn't mean a given object has only one ID;
an object may have many names, or else we have something more powerful
than names, and impossible to implement in a distributed way.
Implementing unicity in presence of communication and computer crashes
means being able to commit on disk, which is hard to do reliably on IDE
disks (SCSI is ok). Unless we want to be extremely slow, it also means
we'll be doing preallocation of IDs, and/or optimistic/speculative
computations with non-committed GUIDs.

> It doesn't seem too important right now, but when the time comes in my
> opinion protocols should be invented on the fly to contain just the
> information necessary.  Maybe a standard meta-protocol to decide on
> special purpose protocols would work.
Yup. This would be a great feature of TUNES over other systems.
Not a feature needed for initial bootstrap, tho.

[ "Faré" | VN: Уng-Vû Bân | Join the TUNES project!   http://www.tunes.org/  ]
[ FR: François-René Rideau | TUNES is a Useful, Nevertheless Expedient System ]
[ Reflection&Cybernethics  | Project for  a Free Reflective  Computing System ]
An apple every eight hours will keep three doctors away.