[Bug 1804] [FS] editorial: E.1.4.2 Erases

http://www.w3.org/Bugs/Public/show_bug.cgi?id=1804





------- Comment #6 from jmdyck@ibiblio.org  2006-04-14 21:54 -------
(In reply to comment #5)
>
> So with the simple (and traditional) definition you have in mind,
> that still results in all simple values also be formal values.

No, I covered that in the first para of Comment #2: the fact that
L(SimpleValue) is a subset of L(Value) doesn't justify saying that every
SimpleValue is a Value. But that's beside the point.

The pertinent question is, what things can legally be bound to an italicized
non-terminal (e.g., SimpleValue) in a judgment?

My answer is: any syntactic object that is actually derived from the
(SimpleValue) non-terminal.

I believe your answer is: anything that could have been derived from the
(SimpleValue) non-terminal, whether or not it actually was so derived. (Perhaps
each "thing" is a member of the language generated by the (SimpleValue)
non-terminal [a series of tokens or characters, which doesn't know how it was
derived], or perhaps each "thing" is an abstract value from a value-space.)

Because Value does not derive SimpleValue (or vice versa), my answer says that
you can't bind a SimpleValue object to a Value italicized word (or vice versa).
Thus, a Value italicized word in one judgment will never unify with a
SimpleValue italicized word in another judgment. (So, e.g., Sem / rule 9 /
conclusion can't match Sem / rule 3 / premise 1, though you presumably want it
to.)

With (what I believe to be) your answer, there are things (strings of tokens,
or abstract values) that can be legally bound to both Value and SimpleValue
italicized words, and so such words can unify, and the judgments containing
them can match.

My problems with that answer are:
-- it's a complication without a counterbalancing benefit, and
-- there's lots of evidence that the inference process does indeed operate on
syntactic objects, not strings of tokens or abstract values.

Received on Friday, 14 April 2006 21:54:21 UTC