Re: MARS last call: packet formats (fwd)

[James Luciani <[e-mail]>]

Yakov,

> > on a single packet format for both NHRP and IP-MC.  I would welcome more
> > comments from the group as to:
> > 
> > a) whether or not this is desireable and
> 
> It certainly would be nice to have this as a goal.
> 
> > b) how we should get to a single format.
> 
> IMHO as a first step we should settle on a single format for the
> Registration message. If other folks think this is a reasonable first
> step, then I'd like to get some feedback on whether it would be
> reasonable to use the IP-MC packet format for registering both
> multicast and unicast addresses.

I disagree.  I do not want to see different packet types
have completely different semantics (i.e., one which has registration semantics
and one or more "other" packet types with there own semantics) within NHRP.

In the extreme, we have 3 distinct choices:

1) we have inconsistent semantics and separate packet formats for both 
   protocols
     -- I think we are all agreeing this is not acceptable

2) we have consistent semantics and separate packet formats for both protocols
     -- I feel that this is an acceptable choice which does not
        require functionality in one protocol to be mirrored in another 
        protocol that would not use it.  Further this approach still allows
        both protocols to benefit from a set of machinery which although
        different in their respective parsing code would nonetheless
        be the same in their processing for that functionality that overlaps.
        This follows the approach I took in my previous notes on the packet 
        formats from a few days/weeks ago.

3) we have consistent semantics and exactly the same packet formats for 
   both protocols
     -- This allows us to have exactly the same code for both protocols.
        While I am not totally against this choice, I think it is 
        too constraining for each protocol.


Recently, there have been a number of suggestions that we do something
akin to a "munge" of 2 and 3 where the semantics are the same and the
general parsing rules follow a set of standardized guidelines (those already 
followed to some degree by IPMC and ATM ARP).  I like this idea as long as it
does not end up either limiting the functionality of NHRP or making the parsing
of NHRP packets harder.  However, I am not sure how much we get out of a munge 
of 2 and 3 over just doing 2.  Also, I would very much like to have this 
discussion put to rest at/before/during the Dallas IETF so that we can work on
any remaining functionality issues.

Regards,
-- Jim Luciani
__________________________________________________________________________
James V. Luciani    Ascom Nexion                    voice: [phone]
[e-mail]   289 Great Rd., Acton MA 01720   FAX: [phone]