IPv6 BGP TE (was Couldflare routing problems)

[Jean-Francois.TremblayING at videotron.com]

> > Out of curiosity do you follow current RIPE PA filtering 
recommendation?
> > http://www.ripe.net/ripe/docs/ripe-532
> > It is suggested that prefix filters allow for prudent subdivision of 
an
> > IPv6 allocation. The operator community will ultimately decide what
> > degree of subdivision is supportable, but the majority of ISPs accept
> > prefixes up to a length of /48 within PA space.
> 
> No we don't, because we do not agree with that reasoning.
> In those rare occasions where it is required to announce more 
> specifics out of /32 globally because of disjoint networks (and 
> apart from that I see no valid reason at all),
> the individual disjoint networks should either get a separate /32 
> each, or get a /48 PI out of the designated netblocks.
> 
> Regards,
> Chris
> AS559

Sorry for pointing at the big pink elephant in the room, but what about
IPv6 BGP traffic engineering? This is probably not a concern on the short 
term, but it could become one fairly rapidly as traffic grows. I 
understand 
the fears about table growth, but there's also a genuine need for ISPs to 
control traffic flows, at least a few ASes away when using multiple links. 


For example, on our side of the pond, using the current ARIN 
recommendations 
of /56s to end-users (https://www.arin.net/policy/nrpm.html#two8), a 
single 
/40 could represent up to 65k residential users. That's probably a few 
gigs
of trafic and would be a good unit for TE. /36 is definitely way too big 
for this purpose. 

Any thoughts/experiences/ideas on this?

/JF
Videotron
AS5769