On 2015-08-05 4:49 PM, Dino Farinacci wrote:
This is very simple to resolve. This is not a technical response.
(1) Have homenet adopt both protocols. Let the market decide. That means there
may be routing protocol redistribution in the home. This would be a mistake.
Would we not want to differentiate the option? Picking both now is not
the same as either being available for use over time (perhaps that's a
version of (2) below with a future option.
I am not sure why we are hung up on the notion that if we pick one now
that it will be impossible to use a different protocol later (either
exclusively or cooperatively). This may not be the most ideal scenario
(changing the options, or default later), but may be the most practical
if we truly think there is future truth which will make one or the other
protocol "seem" better later.
On that point, given I have no real bias either way, picking one (since
it appears that both are viable) seems to be more important then having
the (perceived) best protocol out of the gate - but not making progress
on any seems to be the bigger issue. On my take on perceived, given we
have so little experience on how these protocols fare in the general
case of homenet, many of the points made are quite theoretical (and at
some point we just need real experience).
Many protocols seemed "great" until we actually used them, at which time
we made adjustments to make it work better in real life (meaning real,
non-lab, networks).
Personally, I don't think there will ever be the prefect protocol for
this, just like there is no perfect protocol for any of the networks I
ran (we have chosen, swapped, changed interior routing protocols over
the years in the networks I worked with). The homenet is not likely
going to be the prefect specimen of a network model which will be the
same everywhere. I am sure there will be differences based on many
factors - big one of which is the upstream operator.
>(2) Pick one and move on. This has a chance of working if open source
is available because vendors will take the quick-approach into
>delivering a solution. We saw this work with SNMP in the 90s (but this
wasn’t an open-source case but a single reference implementation case).
> (3) Pick one and hope it catches on. Vendors, if they have vision
will say the chosen one was BS, and they will do their own. That means >
there may be routing protocol redistribution in the home. This would be
a mistake.
As to what may catch on, I would suspect the larger operators will have
big say here. They typically decide on what's on their networks and
what gets pushed to the homes (not in all cases, but in many cases).
Understanding where they are leaning should also be taken into
consideration. Also, if they are leaning in different directions, that's
also important.
So just to put myself in the line of fire now, it appears that we have
had good progress on Babel (yeah I know the WG gap - but that can be
fixed can't it?), but what technological reason is there which makes
just going with this so problematic? Had the case been the same for
IS-IS for this use case profile, I would have stated the reverse.
regards,
Victor K
_______________________________________________
homenet mailing list
homenet@ietf.org
https://www.ietf.org/mailman/listinfo/homenet
