Re: [9fans] UN to fund linux for the 3rd world

[dvd@davidashen.net]

>>>Interestingly enough, a number of projects that were XML-based are now 
>>>S-expression based, so I guess it works for them too. 
> 
> several of us have used non-xml syntax (eg, Xduce) even when one
> of the possible physical representations might be xml.  the scanner
> is tiny, the syntax can be used directly in a programming language without
> itching or retching, and what matters is the underlying abstraction anyway.

How widely is Xduce used?

> xml provides (slightly weird) trees; s-expressions are trees not lists.

Everything is a tree. At the time of SGML, some used to define SGML syntax
to be parentheses based, that is, s-expressions you like so much. But see below.


> all of the non-xml forms i know seem to be clear cut about whether a
> character is or is not part of a string.  i like the precision, myself.

> the power comes from tree operations, not the syntax of the tree
> representation (but see below).

The power comes from both. Not having to think about a parser is a great
relief.

> it would be fair to say that xml's providing a dtd
> was an advance.  

The opposite. The fact XML does NOT need a DTD is one of its most
important advantages, and that is why it only has one lexical syntax
instead of many.

(it was SGML really, but by eliminating `&' XML
> eliminated a real nuisance.)  fortunately, that helpful step was subsequently
> itself eliminated by inventing Xschema, so i don't need to be all that fair!

There is also Relax NG, http://relaxng.org/ .

> but was really a cunningly disguised way of getting people to stop
> using XML directly (let alone XSLT) and start using lisp and s-expressions.

Yes, that's useful sometimes; in my Relax NG (compat syntax) based XML
validator I use embedded Scheme interpreter for extensible data types
library; the library is based on complete implementation of XML Schema
Unicode regular expressions I wrote in r5rs scheme.

Different tools are good for different jobs. But this is not because s-expressions
are better than XML. XML is a tree model and a serialization format; the former
gives tools for validation, the latter gives tools for parsing and generation. Both
work amazingly well; in cases when neither is necessary and a different syntax
is preferrable, for example, for human perception, a different syntax can be
used. 

XML with its data model and serialization format is the best available solution
for most tasks. Where human eyes and brain are involved, parentheses- or
curly-braces-based syntax is frequently used; with equivalent XML representation for the ease of validation and interchange convenience.

XML is like automatic garbage collection: while there are always proponents
of 'better' hand-crafted memory management discplines, automatic gc always
wins because it makes the developer free from the trouble of memory leaks.
At the price of lower performance and greater memory requirements in certain
cases. 

Just make something that works better.

David