RE: Comment on draft-coene-sctp-multihome-04.txt

[Coene Lode <Lode.Coene@siemens.com>]

< Snip...>
You state that:

>     Under the assumption that every IP address will have a different,
>     seperate path towards the remote endpoint, (this is the
>     responsibility of the routing protocols or of manual configuration)
>     , if the transport to one of the IP addresses (= 1 particular path)
>     fails then the traffic can migrate to the other remaining IP address
>     (= other paths) within the SCTP association.

Now in the original context that SCTP was designed for, you might have that 
much control over the routing environment. But it seems to be a very
dangerous
assumption for the general case of two systems communicating over the
whole Internet.

<...snip>

It is indeed true that this particular assumption will NOT be true in the
general case.
The general case would be that along the paths present, some or all of them
will share some routers and/or links between them. SCTP is not able to
detect this and can therefore do nothing to enforce it.

Therefore it is an assumption which can be true: but then things outside the
scope of SCTP have to do something....
Or the assumption can be false: then SCTP will work(just as in the first
case) but then weird things could happen..(example 2 links from host -> only
one is ever carrying traffic, the other one always sits idle due to the
routing table in the host.. See draft: it should be beter explained
there...)
 
<Snip...>

I actually think this points to an important component of any solution
that we haven't really talked about much, except when NAROS was discussed
a few months ago - how does one determine the existence or absence of
connectivity between two locators? For SCTP or any other solution,
it's only when connectivity exists between an address pair that they
are any use.

<...snip>
That is where the heartbeat comes in... Address exchange alone doesn't solve
the problem.
(Heart)Beating the addresses(sory for the pun...) gives you a idea of it
reachability...
 
<Snip...>

By the way, you also say

> 
>     As a practical matter, it is recommended that IP addresses in a
>     multihomed endpoint be assigned IP endpoints from different TLA's to
>     ensure against network failure.

The term "TLA" no longer exists in IPv6 and in any case, it is a false
assumption that two different high-order network prefixes imply
two different paths. They might, but they might not.

  Brian
<....snip>

I better then remove the sentence...
Replacing them by provider seems to limit the choice as multihoming can also
occur using only a single provider..
But it is easier to demonstrate this when talking about 2 or more providers.
2 different prefixes can still take the same outgoing link from the host..
That is not what we really wanted...

Thanks for the comments.. And if there are more, let them come...

Yours sincerely,
Lode