Re: Distributivity in Tropashko's Lattice Algebra
Date: 16 Aug 2005 00:55:54 -0700
Message-ID: <1124178954.235852.266690_at_g14g2000cwa.googlegroups.com>
Let R(a:int, b:int) = {
(5,1),
(8,-1),
(2,3),
(1,111)
}
Then we can calculate f join R
The result is:
(a:int, b:int, root:int) = {
(5,1,2.0),
(8,-1,3.0)
}
Does this make sense? I realize that f isn't strictly "infinite" because it is bound by the implementation of the int and float types, and possibly also by the size of memory, but it is at least no *more* bound than we would have been anyway for a given implementation of int or float. (That is, it introduces no limits the implementation didn't already necessarily have.)
I believe this approach to be mathematically sound and am working towards a notation, and I hope eventually an operational semantics for it. I have been working on it for quite some time; there is some chance it may actually be interesting when I am done. I am quite aware that the odds are against this.
Feedback welcome.
> Besides, I am not sure the function belongs to the NA (new algebra).
It may well not be any part of what Mr. Tropashko is thinking, but
it is certainly part of the algebra that I am trying to construct,
(which has been greatly advanced by his paper.) One goal I have
for this algebra is that it be implementable, so it cannot allow
constructs where we simply wave our hand and join with an infinite
relation, however appropriate that may be for the mathematical
point of view. I am (to my occasional dismay) not much of a
mathematician, but I will claim to be a passable software
engineer.
Marshall Received on Tue Aug 16 2005 - 09:55:54 CEST