[e2e] RE: [Sigtran] Routing for multi-homed host traffic

[Xiaoming Fu]

Hi Lode,

> you can have only a single address for B but then you fall into the case of
> figure 2.1.3(a bit below 2.1.1). With having 2 routing entries in the host
> routing table of A (instead of one), you will be able to use it. And the
> source address can be one of the 2 interfaces, provided you then send it out
> on the interface with that particular soure address. The selection of the
> particular source address must be done by SCTP or the application above. The
> network layer has a choice in selecting any of the 2 links(they both goes to
> 3.2), so use a default (eg always use upper link) or using the source addr
> provided by the transport (and then the network layer has to honour this
> choice by sending it out over that particular link associated with the
> source IP).

Sure, the problem stays, if SCTP or any application above would specify
a source address, the network layer would have to replace the default 
routing selection anyway. 

> A better solution(to me) is to do as in figure 2.1.4 and assign symmetric
> addresses.

Agreed. However it may not work for mobile IP, where a correspondent node 
is expected to use a same IP address while a mobile node moves.

> > http://search.ietf.org/internet-drafts/draft-tkn-nsis-qosbinding-mipv6-00.txt
> 
> Well, then this is a application who can use this. If the SCTP
> implementation does support this , then you can do it.

Thanks. Actually we didn't intend for SCTP in this context, but
I think the same idea could be applicable to extend SCTP in 
mobility scenarios: e.g., establishing an alternative association 
ASAP while a MN (as an SCTP endpoint) reachs overlapping 
area.

Regards,
Xiaoming
> 
> 
> yours sincerly,
> Lode
> 
> Siemens atea
> atealaan 34          2200 Herentals, Belgium
> E-mail: lode.coene at siemens.atea.be
> Tel: +32-14-252081
> Fax: +32-14-253212
> 

--
Xiaoming Fu
fu at ee.tu-berlin.de
http://www-tkn.ee.tu-berlin.de/