[
Lists Home |
Date Index |
Thread Index
]
From: "Rick Jelliffe" <ricko@allette.com.au>
> Executable specs
> for standards (such as IDL) are quite a lot better, but most
> specs are not for interfaces.
Umm ... IDL isn't executable. That's the point. It covers
interface syntax (like BNF) not behavior. Behavior gets
described in natural language. I've made that same
point about needing natural language to describe
behavior. Most people wouldn't know a formal semantic
specification language if it bit them ... :)
Remembering that my point was about "strict interpretation"
of specs needing to be meaningful (else why bother?) in the
specific case of the attribute order issue:
> So the idea of standards as Holy Writ passed down from the
> gods messes everyone up: a standard is the result of negotiations
> from some community, and the best way to make standards
> work is to integrate in with that community and to get to know
> the original intent.
That model of standards doesn't work so well with W3C since
essential parts of the "original intent" are hidden behind closed
doors. And it doesn't work for "remote implementations" where
such "integrating" is for any reason (language? travel? etc) not
practical. Community integration is good, but that's a different
issue from ensuring that base specifications are strictly correct.
Notice I said a "good spec" ... what you're describing is a "loose
specification" model, which among other things only works well
when all players are acting with good intent (or when there's a
monopoly implementation). Such good intent is unusual; more
typically, there are issues where several "original intents" are
in conflict. Good specs identify such cases as "implementation
specific", bad ones leave it up in the air.
More to the point: In your scenario, that community has a clear
responsibility to clearly describe that original intent. That's what
the specification is for, and what specification errata address.
To some extent, additional specifications (XML Infoset in the
case of attributes) can augment the base specification.
- Dave
|
