[ppml] 2005-1 and/or Multi6

[Michael.Dillon at radianz.com]

> > We have to allocate PI addresses to these so-called 
> > end-sites because to do otherwise is restraint of trade.
> > However, it is unwise and imprudent to offer them /48
> > prefixes which are highly wasteful of global router capacity
> 
> I would like to hear from router vendors that they actually do intend
> to do this sort of optimization. 

I don't think that this is relevant to making policy.
For one thing, we cannot be certain which router
vendors to ask since it is entirely possible that a
new vendor will start up at some point in the future
and will include this feature in their implementation
as part of their differentiation from existing products.
And a vendor who says today, that they will not do such
an optimization, could well change their mind tomorrow.
We should not uneccesarily limit the choices of router
vendors.

> The reason I ask is that even if 99%
> of the routes one deals with are short prefixes (e.g., /32) there will
> always be a need to deal with longer ones too. E.g., for routes within
> your AS, etc. So I'm not sure the above arguement actually holds water
> at the end of the day.

The fact that the router needs to deal with some long
prefixes does not prevent the vendor from providing
hardware acceleration for the first 32 bits of
a prefix. We really should not be getting into design
details here. It is enough to realize that longer
prefixes require greater capacity of various router resources.

> One reason why I would not want to give end sites a /32 in this case
> is that at some point in the future, there may well be a need to
> filter prefixes in order to keep the routing system working.

I don't want to give end sites a /32. I want to give a /32
to any network which is multihomed. I don't consider a multihomed
network to be an end site from a network topology point of 
view. I know that many ISPs consider such networks to be end sites
because of their business model, but our policy should be
independent of business models.

> At some point in the future, when filtering is _required_ to keep the
> routing infrastructure functioning, ISPs will want to be able to
> impose filtering that makes at least some sense. Giving everyone equal
> size prefixes pretty much ensures that will never happen, because they
> will all look the same.

If we ever get to that day, then ISPs could charge a fee to carry
entries in their routing table. The filtering (and BGP peering) will
not be done by routers directly but by separate route servers which
can do complex filtering and damping.

--Michael Dillon