H.248: Binary Encoding

[Brian Rosen]

The reason why the ABNF is usually
token=id LBRKT param *(COMMA param) RBRKT
is because the id is what the command operates ON,
whereas the parameters are what the operation DOES.

Add=<termiD>{blah, blah, blah}
or
Stream=<streamID>{...
or
Context=<ContextID{

The bracketed items are the parameters to the command.
All parameters (uniformly now, I believe) are descriptors.

Making the command
Add{<termID>, blah, blah, blah}
Muddies the id and the parameters.

I don't feel all that strongly about this change other than
it's way too late to make it.  It was suggested any time in
the last 4 months we have been dealing with syntax issues,
it could have been examined, weighed, and a judgment call
made on it.  Now, it seems to me that it breaks too many
things we are trying to not change.  People's parsers for
example, the current ABNF (which has a lot of sweat in
reviewers time) for another.

It seems more reasonable to fix the compiler at this late date.

Brian

> -----Original Message-----
> From: Pete Cordell [mailto:pete at tech-know-ware.com]
> Sent: Sunday, October 17, 1999 7:43 AM
> To: ITU-SG16 at mailbag.cps.intel.com
> Subject: Re: H.248: Binary Encoding
>
>
> Hi All,
>
> I'd like to propose a few things for discussion that might
> make the ABNF248
> and ABNF1711 more aligned.  In particular, one difference
> between the two is
> the number of brackets.
>
> ABNF248 has a lot of constructs of the following:
>
>     token EQUAL id LBRKT param *( COMMA param ) RBRKT
>     token LBRKT param *( COMMA param ) RBRKT
>
> and also does (which I suppose is the general case):
>     token [ EQUAL id ] LBRKT param *( COMMA param ) RBRKT
>
> ABNF1711 naturally has the following structure:
>
>     message = *( parameter )
>     parameter = *( tag EQUAL ) value-field *( COMMA value-field )
>     value-field = value / tag / "{" *( parameter ) "}"
>
> This is close to what is in ABNF248 if we move the position
> of LBRKT around
> in the first case:
>
>     token EQUAL LBRKT id param *( COMMA param ) RBRKT
>
> and add an EQUAL in the second case:
>
>     token EQUAL LBRKT param *( COMMA param ) RBRKT
>
> Unfortunately an optional untagged item at the beginning of a
> construct is a
> problem for ABNF1711.  This is because in general (although
> not necessarily
> in the megaco case) it would make the identification of parameters
> ambiguous.  This means that the third variant ideally needs to be:
>
>     token EQUAL LBRKT ( id / "-" ) param *( COMMA param ) RBRKT
>
> This doesn't seem a radical change to ABNF248, and even if
> the code is hand
> crafted ought to be fairly easy to accommodate.  Naturally,
> any use of tools
> should make this even easier.  In my opinion it also looks
> more intuitive,
> although I guess it's a bit boring and lacks a bit of
> designer flair!!!
>
> Is it possible to discuss such changes?
>
> Pete
>
> P.S. The corresponding ASN.1 for the above are:
>
> token ::= SEQUENCE { id AS ? id, param AS ? SEQUENCE OF param }
> token ::= SEQUENCE { param AS ? SEQUENCE OF param }
> token ::= SEQUENCE { id AS ? CHOICE { present AS ? INTEGER, none AS -
> NULL },
>         param AS ? SEQUENCE OF param }
>
> =============================================
> Pete Cordell
> pete at tech-know-ware.com
> =============================================
>