Re: sbnf (striped BNF)

kifer@cs.sunysb.edu (Michael Kifer) writes:
> But why is it necessary to treat a language syntax as if it were some kind
> of data? 

Um, because data is data?    :-)

BNF is a way to talk about formats for serialing data; ASN is a way to
talk somewhat more abstractly, without needing to specify which bits
come first and how they get parsed apart.  Maybe it puts the line
between semantics and syntax in a slightly different place, too, eg with
ordering.

Anyway, no, it's not *necessary*.  I think it's a good idea, and that's
generally the sense I've gotten from the WG as well.   Lots of people
have expressed support for this approach.

> It is not clear what are we gaining by that, but it is clear that
> we had trouble following through with this ambitious program.

It's clear that we've had lots of trouble following through with our
charter.  There are many, many possible reasons for that.  I suggested
SBNF last week as a way to back off a bit from the top-down ASN-centric
approach, to let us think more in BNF terms, but so far it hasn't been
received like that.

In any case, I think I'd want machine-readable data in each dialect
definition which told my software which collections were ordered and
which were unordered.  I haven't tried very hard to make the case for
why I want that.  Process wise, I consider it an open issue.   It
doesn't seem particular hard to provide.

      -- Sandro

Received on Thursday, 6 September 2007 02:43:53 UTC