RE: [More] Session Mobility

[James Kempf <James.Kempf@Sun.COM>]

Amish,

>
>By session mobility I am refering to all possible types of user sessions.  This
>section in the draft certainly does not focus on one type of service, e.g. SIP
>telephony, but applies to all Internet sessions. The draft is also worded such
>that session mobility has to be the responsibility of the network.
>
>I mentioned in my previous email that session mobility "depending on the type 
of
>service" does not need to be the responsibility of the network and multi-homing
>could address some of the session mobility issues.
>
>Managing session mobility on the network would be advantageous for streaming
>delay-sensitive data. Delay tolerant data could be better served by 
multi-homing
>where buffering of data that arrives out-of-order and retransmissions are
>considered acceptable (e.g. http and ftp type sessions).
>
>I think the requirements need to be a little flexible, such that session
>mobility is not only the network's responsibility. Any objections?
>

No objections, but I still don't see a distinction here between:

1) Moving IP traffic, perhaps selectively  (e.g. just http and not ftp), from 
one interface to another on the same device.

2) Moving IP traffic, perhaps selectively, from one device to another.

The first seems fairly straightforward to implement, and discussion about
how is well underway in the IETF. I believe there is even a draft
about how to move selective IP flows from one interface to another.

The second seems fairly difficult to implement, since it requires a case
by case examination of the application protocols. Most application
protocols don't allow changing from one device to another, and I
suspect there will be a fair amount of resistence on the part of
some application protocol groups to providing support for it.

		jak