Re: BoF session in Prague "Formal State Machines"

[Frank Ellermann <nobody at xyzzy.claranet.de>]

Fred Baker wrote:
 
> Having written parsers for table-described state machines, I beg to
> differ. Table-described state machines can in fact be machine-
> readable if they are designed to be. It's just another way to write
> the language.

Makes sense.  But it would be nice to have a common language for such
simple cases, intuitively readable (without manual), and at the same 
time working as input for tools (validators and/or transformations
into other formats).

For something that's complex enough to _need_ SDL putting it into a
plain text 69-columns format is most probably anyway hopeless. 

Any proposal has its own limitations, but that's no reason to offer
no format suited for the publication in RFCs at all.  

> you might not *like* to write programs that recognize ascii-art
> cells and find in them things like input names, new state names,
> conditionals, actions, and side-effects.

Maybe I would like it if a common format using ascii-art tables is
defined.  I won't like it if I have to tweak that program depending
on peculiar features of any individual RFC using this format.

> You might prefer token parsers like lex. But those are matters of
> preference, not possibility.

If the matter of preference starts with "you have perl / xerxes /
..." I'd say "no, not for the O/S on my box".  

Hannes Tschofenig wrote in another article:

| If every working group suddenly has to write state machine 
| documents using the developed language then some folks could
| get a bit nervous.

IMO that can't be the goal, you're also not forced to use ABNF in
cases where it's clearly unsuited.  But if you use some form of
"BNF" please don't roll your own like the horrorshow in RFC 2068,
and stick to RFC 4234 with all minor warts it still has.  Nobody
is forced to use <CHAR>, <CTL>, or the dangerous <LWSP> with its
"lines containing only trailing white space" vs. "empty lines".

Frank




_______________________________________________
Cosmogol mailing list
Cosmogol at ietf.org
https://www1.ietf.org/mailman/listinfo/cosmogol