[CII] One possible scenario

[John Osmon]

On Wed, Dec 02, 2009 at 12:35:33PM -0700, Joe St Sauver wrote:
[...]
> #The advertisements would be /32s, maybe even a /24 at time -- lots
> #of routing slots, but small swaths of space.
> 
> Those more specifics are likely to be ignored by a lot of providers
> who run more-or-less industry standard route filters (e.g., the /32's 
> are particularly likely to be dropped, although even /24's will have
> problems at some providers if the /24 is a deaggregated part of a 
> larger covering netblock). These days people are very concerned about
> the growth in the global routing table, and for very good reasons,
> and the net result is restrictive route filters.

I understand the implications of lots of routing slots -- but I'm
merely throwing out a straw man that could be done with existing 
technologies.  I *want* people to poke holes in the idea.

My next routing policy design is going to have a set of communities for
this on the off chance it is needed.  They may never get used, but at
least they'll be there, and I can attempt to put appropriate language
into peering agreements as well.

After all -- the routes won't ever hit my FIB until they pass prefix
filters *AND* they have the proper community attached.  When I pass
those same prefixes on, the "my net, my rules" filter will allow
people to accept them or not.

Lastly, if the "big boys" are gone, we won't have 300k+ BGP
advertisements to keep in memmory.  So the extra advertisements
aren't likely to cause problems during the critical infrastructure
failure.