Re: addition of TLV to locator ID or locator ID set

[Paul Jakma <paul@clubi.ie>]

On Thu, 22 Sep 2005, Iljitsch van Beijnum wrote:

> I still don't know what you mean here. These days, 85% of the 
> internet is reachable over either two or three hops. So "best" is 
> largely meaningless here.

The general "depth" of the internet may be short in terms of AS hop 
length, that does not mean each AS is equal in terms of actual 
path-length. That one AS path difference might be the difference 
between:

    A_
    | \
    |  Tier-1
    |   |
    Y   X
     \ /
      B

Where A and B are, say, in the same country, both peered with Y, a 
national-level ISP, B also gets transit from national-ISP X, A also 
of some large multinational service provider. Tier-1 however does not 
peer locally, cause that country is just too small for it to be worth 
it, so X-Tier1-A actually goes way out of the way to LINX or AMSIX 
and back.

That's not an uncommon scenario in my country: none of the big 
operators are interested in peering locally. There can be a fairly 
big difference between 2 and 3 AS hop counts - the difference between 
staying reasonably local and going the scenic route via London or 
Amsterdam and incurring a good bit extra latency[1].

> Also note that in BGP we only get to use the "best" path, while the 
> shim allows using different paths in quick succession or at the 
> same time.

This isn't quite true, various BGP implementations are able to 
construct ECMP routes in their Local-RIB and use multiple best paths 
(bit of a nasty hack though). BGP may well acquire better support for 
ECMP in future.

> Testing a large number of paths is exactly the situation we're 
> desperately trying to avoid.

Hmm, well you know my opinion on this.

I really don't think solving routing problems should be within remit 
of this group, other than ensuring shim6 plays nice with routing.

1. Significantly less though than is implied by interleaved DSL, the 
common access technology here, but that's not relevant to other more 
efficient access technologies.

regards,
--
Paul Jakma	paul@clubi.ie	paul@jakma.org	Key ID: 64A2FF6A
Fortune:
Just because your doctor has a name for your condition doesn't mean he
knows what it is.