The Routing Over Large Clouds Mailing List Archive by date

[Mark Laubach <laubach@terra.com21.com>]

>    Folks, the classic RFC1577 LIS model is our default - our point of
>    dependendable always-available fallback.  LISs must be required to support
>    ATMARP by default, or things break in a multi-option, multi-vendor world.
> 
> i strongly disagree with the above.  what a lis is required to support
> is the decision of the guy who runs the lis, not the ietf.  if i
> understand you correctly, even if my lis is all nhrp, i would still need
> to have atmarp code and servers around, which doesn't make any sense.

Juha, let me try to explain it this way.  Even if the world's best and
accepted NHRP appeared tomorrow, my AD's have recently reminded me that
the IAB and IETF interoperability preservation process would require a
step-by-step transition plan to move from what we have now to an all NHRP
LIS.  (I'm assuming we are both defining "NHRP LIS" as the clients only
have NHRP code in them.) The next step of anything we do in a step-by-step
process requires the support of existing implementations and will still
require atmarp code in the client.  Granted, that we decided to aim
towards NHRP at the Danver's meeting - it is just that it takes two steps
to get there, not one. 

Mark