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

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





------- Comment #16 from jmdyck@ibiblio.org  2006-04-18 20:27 -------
(In reply to comment #15)
> 
> I still do not understand what's broken with the current formal semantics spec.

The primary thing that's broken is that the intended interpretation of
"objects" (that can be bound to italicized words) isn't specified.

If (as you appeared to indicate in comment #13), you think the reader is free
to choose an interpretation ("this does not enforce any specific way to
represent the syntactic objects"), I refer you to the conflicting results of
comment #6.

If, on the other hand, you think that the reader is *not* free to choose an
interpretation, and must use the one that you favour, then I would argue that
it's a poor choice. Under that interpretation, when you write a judgment, it's
difficult to be certain that it only applies when you want it to, since it
might unintentionally apply to some string of terminals that happens to be in
the language generated by one of the judgment's nonterminals, even though those
tokens aren't actually derived from that nonterminal in the syntax tree. And
it's similarly difficult for me to determine whether a judgment applies when
you don't want it to. So I can't currently point to a rule that would break
under your interpretation, but that's because it's prone to subtle,
difficult-to-find breakage.

If on the third hand, the FS embraces the interpretation I favour, then E.1.4.2
/ Sem / rule 9 is broken, and probably a few others. (And I would work on ways
to fix them.)

-- 

> I think the interpretation used is pretty clear.

Given this conversation, I don't understand how you can think that. Can you
indicate a place where you think it's clearly conveyed? (The only text you've
quoted is from 2.1.2, and you have my response to that.)

> I do not believe there is any indication that non-terminals are being
> part of the various objects being processed through inference rules,

And I believe that use of the phrase "syntax tree" *is* such an indication.

> and I think that interpretation is used consistently throughout the document.

That's difficult to disprove, since it's a much less tractable interpretation.

Received on Tuesday, 18 April 2006 20:28:03 UTC