xml-dev - Re: Request for help understanding the XML Recommendation

[Rick JELLIFFE <ricko@geotempo.com>]

"DuCharme, Robert" wrote:
 
>  My old CS textbooks aren't in the office where I'm typing this, so I can't
> double-check whether multiple start symbols are allowed in a formal grammar,
> but to me, the situation described above points to extSubset having start
> symbol status itself. This is how it can play a necessary role in the spec
> without being referenced by the right hand side of any other production.

There is a long-running difference in expectations between those who see 
productions as defining the language and those who see productions as 
a formality which helps specify the language. The first lot will
interpret
anything unusual as deficiencies in the specification; the second lot
will
interpret anything unusual as deficiencies in the expressiveness of the
formality.

Of course, it is not impossible that there is some formalism that can
completely describe XML (Bert Bos has an attempt using ADT somewhere). 
But then the problem exists that if people are not familiar with the
formalism, there is little point in using it for the initial
specifications of a language that has a goal that implementers can
quickly produce a rough parser. 

So there is a difference in expectations about whether the text or the
productions are the cart or the horse, and in any case simple grammars
are not powerful enough to be used as a horse to describe the class of
languages we want anyway.  Indeed, this is the long-running problem
faced by compiler compiler creators: how to they find the best
formalisms for expressing non-toy languages.  

Rick Jelliffe

***************************************************************************
This is xml-dev, the mailing list for XML developers.
To unsubscribe, mailto:majordomo@xml.org&BODY=unsubscribe%20xml-dev
List archives are available at http://xml.org/archives/xml-dev/
***************************************************************************