Scott,
Why "stack" instead of "node"?
We need node identifiers and node locators, identifiers mapping to locators.
On Monday 13 July 2009 21:02:08 Scott Brim wrote:
> Toni Stoev allegedly wrote on 07/11/2009 1:21 AM:
> > Please, anyone who writes on identifiers topic, identify your empathy to:
> > Design Goals for Scalable Internet Routing, 3.6. Decoupling location and
> > identification
> >
> > I do care about that goal.
>
> Of course.
>
> > We have interface (node) identification with IP addresses.
> > It is used for node-local session identification as part of the socket.
> > There the IP address is an identifier of the endpoint of the inter-node
> > communication process.
>
> Please let me know if I am paraphrasing you correctly:
>
> - an IP address identifies an interface
> - identifying an interface is the same as identifying a node
> - an IP address is part of the identification of a session endpoint
An IP address identifies an interface and thus generally the node that it
belongs to.
This is the IP identification of a node.
> IMHO an IP address can be part of the set of identifiers you use for
> _initial_ contact, to establish a session in the first place (this is
> the "stack ID" discussion). However, you want a session to be able to
> run over more than one interface, so you quickly want to decouple
> identifiers used for session continuity from any interface identifiers.
>
> > Let us extract this identification from IP addresses and have it separately.
> > But let us keep using the hierarchical structure for the function of
> > authentication of identity.
>
> If I understand correctly, you are saying that the set of {IP address,
> port, protocol} can be used as input to authentication. For
> authentication you generally need a trusted third party and identifiers
> that third party understands.
And identifier structure that points third/intermediate parties.
> Toni Stoev allegedly wrote on 07/11/2009 2:16 AM:
> > Scott Brim wrote:
> >> If I move a session from one device to another, or from one
> >> interface to another,
> >
> > Scott, a session is established with a node. You can move
> > communication among interfaces but you keep talking to the same node.
>
> A session is established between communicating entities; those entities
> may move from node to node. Don't forget about virtualization.
Are those entities application instances or do you mean virtual nodes?
> >> a node identifier is not enough to support session continuity.
> >
> > A node locator, a node identifier and a node-local session instance
> > number would do the job.
>
> You don't want identifiers used for session continuity to be required to
> change just because a node changes its identifiers. The local session
> instance identifier is the only one of these that may work for you.
Again, what are the session endpoint entities that communicate?
> >> If I have a way to identify sessions that does support session
> >> continuity, then a node identifier is not necessary in addition for
> >> that purpose.
> >
> > A node locator and a node-local session instance number are
> > sufficient to have a session. A node identifier is needed to move the
> > session from one node to another.
>
> Now I'm confused.
Sorry about that.
A node identifier is needed to move the session from one node location to
another.
Toni
_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg
