[Battlemesh] Session about multi-homed IPv6 mesh networks and auto-configuration

[Ferry Huberts]

On 19/04/16 19:10, Juliusz Chroboczek wrote:
>>> You could either put the information in an optional Babel sub-TLV, or in
>>> an optional HNCP TLV.  The allocation rules are similar for both
>>> protocols: either use an experimental TLV number, or document the
>>> extension and get yourself an official TLV number.  (HNCP requires
>>> a proper RFC, while Babel only requires a published specification to be
>>> reviewed by an expert chosen by the IETF, which is much easier.)
>>>
>>> I suggest having an implementation first, though.  I'll be happy to help.
>
>> OLSRv2 has the concept that "announce prefixes" have a metric cost
>> attached...
>
> So does Babel.  But Ferry wanted to announce the raw throughput, not to
> encode it within a metric.


I didn't say that.
I said "We really need something like that ..."

:-)


My point is that in order to take an informed decision about which exit 
link to use we need _something_ that correlates to how 'good' the exit 
link is.

This can be raw bandwidth, latency, load, whatever. Either individually, 
mixed together in some kind of abstracted 'metric', whatever. This 
probably needs lots of discussion.

My personal preference would be to have all of these properties of the 
exit links available to clients, but I doubt whether that is practical.

Clients really do need more information about exit links than just their 
existence in order to make an informed decision.

If we can get to a point where that is the case I think we'd have 
improved the world a bit ;-)