Hi Bill,

|> |> / RFC 1887 4.4.2 is obviously not functional with *currently
|deployed
|> |> / network stacks*. It would need the kind of software described in
|> |> / strategy B or H in order to be viable. Or do you disagree?
|> |>
|> |> It depends.
|> |
|> |If it's conditionally functional and the conditions aren't met then
|> |it's not functional. Right?
|>
|> Not agreed at all!!
|> Why block a good solution for certain use cases, because there are
|(few?)
|> other use cases that cannot use it?
|> Why force an imperfect solution that has (strong?) disadvantages for
|some
|> (few?) use cases?
|
|Teco,
|
|I think you might be missing my point. I wasn't making a value
|judgment on the reasonableness of using multiple addresses in order to
|multihome. I was simply observing that the users have spoken and what
|they've said is that the currently available tools and protocols are
|unacceptable.
|
|Multiple-PA multihoming is conditional on a number of factors which,
|in current practice, aren't true. So if we want a multiple-PA solution
|like strategy B or H, its necessary to rearrange things so that those
|conditions -are- true. Saying "go use it" when there's not an adequate
|toolset to go use is really no different than the "do nothing"
|strategy.

I say "go on with strategy B / H".
I am not saying users should have used PA multi-homing, even if there was no
adequate toolset.
If something is missing, we can fix it.


|Anyway, if you really want to rehash the "why's" in detail, that's a
|subject for another thread.

I just want the "Summary of architectural solution space" to be neutral and
to be based on facts.


|> By the way, the term SID could be somewhat misleading. Lets
|distinguish
|> transport layer ID (connection ID) from session ID.
|
|Would you expand on that? I'm thinking of a session as an ID that
|conceptually speaking, lasts from the TCP SYN to the TCP FIN. Is there
|a better definition of session? What's the definition of transport
|layer ID?

For transport protocols, the correct term is connection.
A TCP connection is identified by a pair of sockets (RFC793).
One could say UDP is connectionless.

In the TCP/IP model (e.g. RFC1122), there is no concept of a _session layer_
nor _session_.

Although there are IETF protocols that act on a session layer (e.g. SIP),
for discussion on TCP (and even UDP) I prefer the term connection. Use CID
for this?

(Shim6 uses ULID: Upper Layer ID).
(SIP: Recently, there is I-D on Session-ID posted, with xref to UUID
(RFC4122). In this proposal, IP addresses are NOT used for Session_IDs)


Teco.


_______________________________________________
rrg mailing list
[email protected]
https://www.irtf.org/mailman/listinfo/rrg

Reply via email to