Re: [v6ops] PI [ULA draft revision #2 Regarding isolated networks]

[John Curran <[e-mail]>]

On May 30, 2014, at 10:37 AM, Nick Hilliard <[e-mail]> wrote:

> On 30/05/2014 15:21, John Curran wrote:
>> On May 30, 2014, at 10:13 AM, Nick Hilliard <[e-mail]> wrote:
>> 
>>> On 30/05/2014 15:09, Nick Hilliard wrote:
>>>> e.g. asr9000 typhoon cards. 
>>> 
>>> correction: typhoon cards have 4M FIB "credits".  ipv4 entries take one
>>> credit; ipv6 take two.  So assuming no ipv4, a typhoon card will handle 2m
>>> ipv6 FIB entries.
>> 
>> Acceptable end state does not imply viable intermediate states.
> 
> itym, "acceptable intermediate states do not imply viable end state".

No, I mean that the fib capacity _assuming no IPv4_ (end state) is not 
particularly relevant, since most folks have to deal with IPv4 and IPv6 
at the same time.  While Geoff Huston's graphs show fairly predicable
historic IPv4 routing growth, we also have not experienced (yet) the 
significant pressure to route smaller and smaller IPv4 remnants that 
is likely to occur as both regional and service provider free pools
begin to run out...

> We can all agree to that, but we have 60x overhead on current generation
> technology and linear growth.  There are more pressing problems than this.
> Let's deal with them instead.

I'm not suggesting that there is a imminent crisis (or that any particular
work should be done/not be done with respect to ULA's), but was simply noting
that folks should not presume our current trajectory to be safe and assured, 
particularly based on the ability of the latest hardware to cope with an
IPv6-only world.

FYI,
/John