Subject: Re: Better Dylan syntax?
From: Erik Naggum <clerik@naggum.no>
Date: 1997/09/07
Newsgroups: comp.lang.dylan,comp.lang.lisp
Message-ID: <3082654357249434@naggum.no>


* Paul Prescod
| It's a little redundant to write a parser anyhow. Maybe it would be
| better if the compiler's parser were provided as a library that you could
| use from your favourite macro language. This is, for example, how people
| typically work with SGML -- through an API or pre-parsed "pablum"
| form. It seems to work, but people still complain that parsing SGML is
| too hard so obviously they want to write their own parsers for whatever
| reason (habit?).

there is no agreement on what an SGML document "means", and widely varying
ideas about what should and should not be _unavailable_ after the parsing
has completed.  this means people _will_ want to handle the raw text form
of SGML as opposed to some arbitrarily chosen non-textual form.  there is
some consensus in the SGML community that two SGML documents are the same
only if their textual forms are bitwise identical.  there are many attempts
to use _comments_ for various half-witted purposes, and comments are
typically ignored in the parsed forms.  etc.

had SGML defined a "read syntax" and talked about objects at a post-parse
level, SGML would have been a lot more interesting.  in SGML today, every
whitespace is sacred, every inferred delimiter must be marked as not
present in the text, etc.  you just can't handle this mess intelligently by
defining a _single_ "API", and you will always find people who are smarter
than the last guy who defined a "complete API".  it's sad, really.

#\Erik
-- 
404 You're better off without that file.  Trust me.