Re: Review comments on draft-ietf-shim6-proto-03.txt

[Geoff Huston <gih@apnic.net>]

this illustrates the difference between explicit locator preferences and 
the lack of locator prefs.

If the ULP (application actually) has a demand for particular path 
characteristics then, in the context of this base specification, the 
interaction between the  SHIM6 and the ULP would be for the ULP to request 
a forked context and then request the active locator set from the SHIM, and 
then (by setting a NO SHIM flag in the API) probe the locator pairs and 
discover their path  characteristics according to the ULP's preferences - 
then communicate the relative preferences back tot he SHIM using an 
explicit preference signalling to the SHIM.

But my comment was, as I noted in the mail to Spencer - in the _absence_ of 
any preference indicators, then is the ULID a "preferred" locator over 
other locators?

Geoff


At 01:50 AM 17/01/2006, Pierre Baume wrote:
> Hi Spencer,
>
>   Sorry for being vague. Better wording is welcome.
>
>   A combination of the following criteria (or more) could be used to
> choose the best ULID pair.
>
> - Processing time,
> - Bandwidth utilisation,
> - Financial cost,
> - Latency,
> - Reliability,
> - Etc.
>
>   My point was that, after a failure is fixed, the initial ULID pair
> might not be the best pair anymore.
>
> Pierre.
>
>
> On 1/16/06, Spencer Dawkins <spencer@mcsr-labs.org> wrote:
> >
> > I'm not sure what the "costs" are in this statement. Are we just saying 
> that
> > we're balancing the processing time to figure out that we no longer have to
> > encapsulate against the encapsulation overhead? or are you thinking of
> > something else?
> >
> > Spencer
> >
> > ----- Original Message -----
> > From: Pierre Baume
> > To: Geoff Huston
> > Cc: shim6@psg.com ; erik.nordmark@sun.com ; marcelo@it.uc3m.es
> > Sent: Monday, January 16, 2006 2:55 AM
> > Subject: Re: Review comments on draft-ietf-shim6-proto-03.txt
> >
> > Hi Geoff and all,
> >
> > On 1/16/06, Geoff Huston <gih@apnic.net> wrote:
> > [...]
> >
> > > 9 - section 11 - Sending ULP payloads
> > >
> > > As a meta consideration here, is there any logical reason to prefer the
> > > initial ULIDs as locators?
> >
> >   For what it's worth, I don't think that this belongs within the protocol.
> > The protocol should make it possible (maybe via a negotiation), but the
> > decision needs to be left to the hosts (and their parameters/options). They
> > might want it in some cases and in some cases not. It depends on the 
> 'costs'
> > associated with the ULID pairs.
> >
> > Pierre.
> >