On Jul 13, 2009, at 11:47 AM, Ted Hardie <[email protected]> wrote:
At 7:37 AM -0700 7/13/09, Eric Rescorla wrote:
This is a category error: we don't depend on the Update message to
determine
whether an attached node is a client or peer because client isn't
an important concept in RELOAD. Rather, the rules cause peers to
treat
adjacencies
correctly whether they are peers, clients, or peers who haven't
completed the join
process yet. There is no need for a peer to detect or remember
whether an
adjacency is a client or peer.
-Ekr
Speaking personally, I think there is a fair chance that a peer
might want
to treat an adjacent node differently if it is a client and if it is
a peer that intends
to but hasn't completed the join process yet. I understand the
"a stranger is a friend you haven't met" theory behind the current
approach
which makes them equivalent. But that understanding is based
largely on
the explanation given on the list. I think that some small
promotion in
the concept of client from our current stance would be useful.
We already know that people will do this, and having explicit text
on how
strikes me as a good idea.
As I said before I'd be okay with just text in the document that
expounded the current theory. That would help later readers better
understand
why long-term adjacencies without complete join processes shouldn't be
treated as errors. My personal style preference would be for a more
explicit set of distinguishing characteristics, but I recognize that
may be
largely a style thing.
I have added additional explanatory material in the forthcoming draft
Ekr
_______________________________________________
P2PSIP mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/p2psip
