[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: draft-kurtis-multihoming-longprefix comments



There is two sides to this

a) Why do people multihome?
b) Why do they want PI-like address space?

What worries me is if people are doing a to get to b. I don't want to
see these two being the same.
People multihome because it makes their connection to the net more
reliable, because it is sometimes cheaper and because you get better
performance if interconnection between ISPs isn't very good.

People want PI space because that way changing ISPs is trivial. And even
for people who don't want to multihome this is very important: if your
ISP goes belly up, you want to be able to connect to a new one very
quickly.
Well, this is what I intuitively would say as well. However, there is (as you yourself noted in an earlier email) not that many people that multihome. There are a lot more that get PI space. I am still doubtful that we understand the issue.

I am leaning to the conclusion that what you want depend on what you are. And each will require a different solution.

In principle, a "advanced" DSL home user want to multihome for price. Here are multiple address solution might be the right answer.

If I am a SOHO, I am probably also after cost benefits.

If I am a on-line broker I am probably after resilience, and a routing solution might be what I want.

If I am IBM, I will fall under any provider category.

I think something that is very important to keep in mind is what Tony Li said, we need to get away from the model where when a user want's to multihome he will add a entry in the DFZ. The question is, does that apply for everyone and do we think we can get there in one single large evolutionary step?

I think we need to get there; we will get there; but we will have to do it bit by bit.


In theory, renumbering isn't a big problem: you only need to change a
DHCP configuration or some prefix advertisements. In practice, it is a
nightmare because IP addresses show up in way too many places, often to
enforce access restrictions. Getting rid of this is probably the hardest
part in any multiple-address solution.
I know of some really large re-numberings. I think most of the time, people are making a larger fuss of this than it is.


Best regards,

- kurtis -