Re: recent changes requested on the list

[Eric Gray <[e-mail]>]

Jim,

	In answering your question(s) I may use certain terms that are either unfamiliar
to many of the rolc listeners or have no direct analogue in rolc/NHRP.
I ask you, and other listeners (Andrew) to please bear with me on any missuses
that don't obscure meaning too much...

	With respect to 'A' - what we actually need is a 'reserved for MPOA use' bit in
the flags field; how and why we intend to use it should not be a concern within
rolc. In the event that there is anyone who was involved in the generation of the
request, the 'source type' bit is just a little to coarse for what we need in
MPOA.  In the distributed routing model currently in use for MPOA, we need to
distinguish the 'fingers' of the router from the 'palm'; special handling is
required to ensure that responses are associated with 'edge devices' (EDFGs) that
originate the request via the same 'route server' (RSFGs) to which they were
originally sent - because the extra-cloud entities for which the request was made
might be reachable via more than one 'edge device' (and consequently, NBMA
address).  While this distinction can be made by treating these 'edge devices' as
hosts, they also need to be distinguishable from cloud-local hosts (AHFGs) which
can only be reached via one NBMA address.  I am not sure why this is not also a
problem/concern for NHRP, but others who have been following rolc discussions
longer than I tell me that it is not.

	With respect to 'B' - we (MPOA) suggested this as a per-address field because of
the possibility that a protocol may (one day) be defined that allows for variable
length addresses.  Not specifying a length on a per address basis would mean
specifying both a length long enough to hold the longest such address and a
'justification' and fill approach for addresses not this long.



							Eric Gray

James Luciani wrote:
> 
> Folks,
> 
...
>   Things I did not address below are:
>   A) EDFG Origin Flag requested by Eric Grey
>   B) Yakov's and Bruce's request to add internetwork layer address length
>      fields on a per address basis rather than one per packet.
>   C) Russel's request to add a destination address mask field to the
>      IP-specific part of the Purge packet.
> 
> Since I am writing this, I get first shot a commenting on A, B, C :-)
> 
> I have no conceptual problem with "A" but I am not completely clear why
> this is needed but perhaps I should just read the updated MPOA doc.
> 
> I'd like more info on 'B'.
> 
> On 'C', I'd prefer to just allow the destination prefix extension to be included
> in the purge packet (which it currently is not) and I would very much prefer
> to see this be optional and not a requirement.  I do see value in it though.
> 
...