On Wed, 13 Aug 2003, Markus Hofmann wrote:
Alex Rousskov wrote:
Yes (assuming there is not one available already). Please note that
proposed IRML is also a new domain specific language.
Yup, but the usage of XML, for example, allows to very easily and very
quickly implement an IRML parser using exisitng XML parsers - you
almost get it for free.
I believe this is a common misconception about XML. XML does not make
parsing much easier than any BNF-defined syntax. Just like there are
existing parsers for XML, there are existing parser-generators for
arbitrary BNFs (YACC, Spirit, etc.). Practical usefulness of either
for production interpreters and compilers is about the same.
Sure, if there are other (or even more approriate) existing
schemas/languages/etc., we need to discuss their suitability and
decide on the most approriate one. All I'm saying is let's have a
look at what's out there rather than always wanting to design our
own new stuff from scratch... which I believe to understand you
agree with.
Absolutely. XML parsers exist. BNF parser-generators exist. IRML
defines an XML DTD. It could use a BNF instead. Both approaches would
reuse existing facilities to a similar degree.
Correct. Using XML was partially inspired by CPL (Call Processing
Language), which is used for similar problems in the SIP area. We
should learn from their experience - both positive and negative.
Interesting. I need to look CPL up. My comments are based, in part, on
experience with ACLs (Access Control Lists) languages and other
configuration issues for HTTP intermediaries and L2-7 switches.
Thanks,
Alex.