[SCL] The semantics of "semantics"

[John F. Sowa]

Chris, Tanel, Jim, et al.,

I believe that somee recent lapses in communication have resulted
from undefined notions of what "semantics" means.  Those of us who
have been working with logic for some time naturally equate that
term with "model-theoretic semantics" as defined by Ockham, Peirce,
Tarski, Kripke, Montague, etc.

An "impedence mismatch" has come to the fore in the exchange
of notes between Chris and Tanel and my recent exchange of notes
with Jim, which ended in a proposal for some future dinner party,
in which they would be on the menu.

To clarify the issues and the air, I would like to apologize for
any aspersions I may have cast upon the beauty of RDF syntax or the
semantics of the semantic web.  To avoid further misunderstandings,
I suggest the following guidelines:

  1. We should avoid using the word "semantics".  Instead, we can
     use phrases such as "truth value" or "model theory" and talk
     about "English definitions" or "formal definitions", where
     "formal" implies the use of some version of logic.

  2. Discussions about syntax should be raised above the level
     of sniping criticisms about beauty or about the widespread
     usage of one notation or another.  Instead, we should emphasize
     that the SCL model theory is stated in terms of an abstract
     syntax, which can be expressed in various concrete notations,
     such as KIF, CGIF, and many variations of XML-based notations,
     including RDF.

  3. Questions about how to do the mapping from one syntax to
     another are appropriate, but gratuitous remarks about the
     desirability of any particular syntax should be avoided.

  4. Compatibility with the W3C conventions for naming and
     linking is a non-negotiable requirement for any and every
     concrete syntax used to express SCL.  That requirement can
     be most easily satisfied by providing support for the W3C
     conventions directly in the abstract sytnax of SCL.

  5. Since RDF has been designed by the W3C, it serves as a
     critical test case for the claim that SCL is general
     enough to specify the W3C requirements.  Therefore, we
     should demonstrate two mappings:  first, that all of RDF can
     be mapped to SCL, and second, that some appropriate subset
     of SCL can be mapped to RDF.

These guidelines won't prevent disagreements, but I hope that
they can focus the discussion on the important issues.

John Sowa