- From: Tim Bagot <tsb-w3-style-0005@earth.li>
- Date: Sun, 5 Jan 2003 14:06:57 +0000 (UTC)
- To: <www-style@w3.org>

At 2003-01-05T01:58-0600, Shelby Moore wrote:- > At 10:00 PM 1/4/2003 -0800, Sandy Moss wrote: > >1. Tim Berners-Lee has stated, in no uncertain terms, > >that the semantics of a specification should not and > >indeed _cannot_ be orthgonal to the implementation. > > > Of course this is true and I emphatically agree. This is what I have been > arguing for. Thanks for supporting my argument. > > Semantics is controlled by implementation (not orthogonal). I agree with that. That's not what the quoted text means. Non-orthogonality implies that the two things are not entirely independent, and no more. The above would be satisfied, for example, by "implementation is influenced by semantics", which should clearly be true. > [...] > >2. If we assume for the sake of argument that > >semantics are _not_ completely controlled by > >specifications, > > then user agents would have at least > >partial normative control. This would probably lead > >to the orthogonal (if not direct) deterioration of the > >web, which is equally inconceivable. I have no idea what "orthogonal deterioration" might mean. Please could Sandy elaborate? > Note an error in your logic. If semantics is not orthogonal to implementation > (as you assert above and I agree), and specification is not orthogonal to > semantics (per you implication that normative _control_ in UAs is > "inconceivable"), As long as the specification supplies some semantics, the two will not be orthogonal. > then this would mean that specification is not orthogonal to > implementation. The conclusion is true, but does not follow from the two stated postulates. In fact it still does not follow if you delete any or all of the three instances of "not". > We may desire for specification to be non-orthogonal to implementation, but the Certainly. > reality is that implementation non-conformance exists. Thus it follows that But this does not imply orthogonality of implementation and specification, only (at most) the possibility of orthogonality. An "implementation" that completely ignored the specification would almost certainly be both non-conforming and useless. > semantics is (_some_ times) orthogonal to specification. That is essentially > my Axiomatic Proof[2]. Semantics can be orthogonal to specification only when the specificaton supplies no semantics. This is certainly possible, but is not the case with HTML. Tim Bagot

Received on Sunday, 5 January 2003 09:07:01 UTC