Lists Home |
Date Index |
From: "Frank Richards" <email@example.com>
> Uh, wrong. This is not in general government contracting. If the people who
> have to implement specs can't understand them, the specs won't be
> implemented. And having to get help from the specifiers here on xml-dev goes
> against the whole idea of using an ISO spec in the first place: I and any
> other geek in the world should be able to figure out how to meet a spec, and
> whether an implementation meets that spec, without needing personal guidance
> from the author or a separate (and specific) book from Oxford University
> In my previous life as an electrical engineer I had occasion to read many
> specs from CCITT, EIA and the IEEE. (Admittedly none from ISO) Until the
> 8879 set of specs rises to the standard of comprehensibility set by these
> organizations, rather than falling to the standard set by the USDOD
> (actually below it -- you can parse a mil spec if you try hard enough. 8879
> cannot be done without a gloss) those specs will, if not fail, at least be
> severely handicapped in the marketplace of ideas.
I cannot quite figure out whether Frank has read ISO 8879 or not ("none from ISO").
But ISO 8879 is actually a fairly well-written spec in the scheme of things.
I used to use it first, then read Goldfarb's annotated version if the standard
was not clear. The problem with it is that the notations that computer
science provides are not up to the job of defining SGML clearly (the
lack of functions in BNF, to express macro inclusion, for example). So
instead there are informal tables and text in places that cry out for some
more formal expression.
There may be better formalisms now (perhaps in a functional language?)
that could help reformulate ISO 8879, to help parser-writers. It would be
an interesting and useful project.
One of the other problems with ISO 8879 IMHO is that, following
James Clark's sgmls and SP, I think many people expected that
in order to implement a conforming SGML system you had to support
lots of the optional features. (You only have to support the minimal
Reference Concrete Syntax, actually.)
Personally, I think that in about 5 years time (when XML 1.0 is superceded?)
that it will be a good chance to revise ISO 8879 with a thorough rewrite,
re-factoring it for the current fashions.