[Date Prev][Date Next][Thread Prev][Thread Next][Author Index][Date Index][Thread Index]
version compare and partial orgls
- To: <tribble>
- Subject: version compare and partial orgls
- From: Mark S. Miller <mark>
- Date: Fri, 1 Sep 89 17:37:14 PDT
- Cc: <us>
- In-reply-to: <Eric>,00 PDT <8909012321.AA05157@xxxxxxxxxx>
Date: Fri, 1 Sep 89 16:21:00 PDT
From: tribble (Eric Dean Tribble)
The unify operation does not (and should not) require that the orgl be
partial. Two version of hamlet may be complete, but they can be
unified with each other (assuming they are equal).
Hmmm, interesting. You're right, in that the sense in which a
unifiable orgl is partial is separable from the sense in which an
informable orgl is partial. I have been assuming that this
distinction isn't made; it isn't clear to me whether or not the
distinction should be made.
Note that the two versions of hamlet are both represented (probably)
by MapOrgls which map to DataStamps. These MapOrgls may be complete
even if they map to DataStamps which are informable and/or unifiable.
The way one unifies one version of Hamlet *to* another (it is
important to always use assymetric terminology when speaking of unify)
is by unifying the DataStamp(s) that that version of Hamlet maps to to
a corresponding mapping of the other DataStamp.
Now an EAddress (StampID * iD, Position * pos) can be interpreted as
saying: "I don't know what its true identity is, but its identity is
the same as the object at pos in the Stamp identified by iD.". If
this Stamp is a unifiable DataStamp, then that is the best we can do
wrt identity in our current state of ignorance.
Inform informs the DataStamp of more things that live in it, whereas
unify (which turn it into a MapStamp) informs it of better identity
information than it had before. Both are monotonic increases of
knowledge, that only those who have grabbed the DataStamp with a Foot
may tell it. Now the question it seems to me comes down to: "When we
want to let a DataStamp know that it will never get either more
informs, or ever be unified, are these two separate actions?" Or
equivalently: "Does a DataStamp have two state flags: 1) whether or
not it is complete, and 2) whether it is unifiable or frozen; or does
it only have one flag which turns both of these abilities off
In either case, new information added to the DataStamp can show up in
already existing MapOrgls which map to it.