Dave,

Dave Thaler wrote:
> > How does the application requesting a particular lifetime effect the
> > lifetime that the routers use when advertising the unicast prefix?
> 
> It doesn't.
> In the abstract API, the application supplies a minimum and desired
> lifetime.  If the app says [0, 6 months] and the host knows that
> the unicast prefix is valid for 1 week, it can tell the app it has
> it for one week (and the app can renew it at any time).

That was my intent.  It fits the existing v4 mcast allocation
architecture.  A host may request an address for ANY particular
lifetime, but there are no guarantees that the requested lifetime
can be honored.

> 
> > Are you envisioning some new protocol to inform the routers (and the
> > system
> > admin, ISP, RIR) that allocated the unicast prefix that it needs to
> > stay around for long enough to satisfy the applications request? :-)
> 
> Gosh I hope not :)

Agreed!  I am trying to reduce the number of protocols.

> 
> > So while there might be an abstract API that allows the application
> > to ask, the scheme in uni-based-mcast doesn't seem to have the pieces
> > so that the system can honor such a request; at least not with a
> strict
> > interpretation of the relationship between the RFC 2462 notion of
> lifetime
> > of the unicast prefix, and the presumed lifetime of the derived
> multicast'
> > address.
> 
> If the app says [6 months, 6 months] then personally I would expect the
> host to fail the request if it only has a unicast prefix lifetime of 1
> week.
> I think your question is really: should we allow such a host to grant
> a request for 6 months if it only has a unicast prefix lifetime of 1
> week?
> 
> In my previous email, if there is a need to allow this (and I'm not
> saying
> there is), then we'd have to just say it SHOULD NOT, and say that after
> the unicast lifetime, the multicast packets are not guaranteed to be
> routed.

I have no problems changin the text to SHOULD NOT.

> 
> > Hence my question whether the intent is to have such a strict
> relationship
> > or not.
> 
> I'd like to hear others opinions, but I think that is the authors'
> intent.

It was my intent.  IF we are going to try and eliminate the multi-
layered allocation architecture, there needs to be some controlling
mechanism and the only one I could find was the unicast prefix in
the RAs.

Brian

--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page:                      http://playground.sun.com/ipng
FTP archive:                      ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------

Reply via email to