Sebastian and all, I'm late for answering your questions.
I would answer "yes" to the first two questions you raised. For the third question, I think the discovery document is not going to standardize the way for ALTO server discovery, it is only standardize the service tag for one discovery option using U-NAPTR. There are variable discovery methods available. So there is no one standard way. Or could you specify one standard way that every application should obey to achieve interoperability? > If you use triangular routing it might make sense to use the > care-of-address instead of the home address. But this may depend on many > things, e.g., are you mostly sending data or mostly receiving data? > So it may also be wise not to do that ... > This is a good question. It should use one ALTO server when it receives data and another ALTO server while it sends data. Should one end-host be tied with only one ALTO server? It is easy to generalize the mobile and proxy based scenarios I raised, or even generalize the whole document into only one short paragraph such like "find the resource consumer's domain name, then use the U-NAPTR method with the new defined service tag...", if we do not go to the details. BR, -Haibin > -----Original Message----- > From: [email protected] [mailto:[email protected]] On Behalf Of > Sebastian Kiesel > Sent: Sunday, July 01, 2012 2:49 AM > To: Songhaibin > Cc: [email protected] > Subject: Re: [alto] ALTO discovery in roaming scenario > > Haibin, all, > > On Sat, Jun 30, 2012 at 02:35:06AM +0000, Songhaibin wrote: > > I'm very glad that you understand my proposal. If the text missed some > > technical details of ALTO discovery in mobile IPv4 and mobile IPv6, I > > think we can clarify them in the draft. But I cannot agree that we do > > not need text about the right IP address selection (in mobility and > > proxy based scenarios) in the document. > > > > What do other people think on this issue? > > I like Michael's proposal ... > > > > "In general, the ALTO server discovery should be based on the IP > > > address that is used to communicate with other peers". > > ... because it is short and generic. I think it is better to give > general guidelines (if possible) instead of going into the details > of various mobility mechanisms. > > > I think Michael's proposal is perfect for all mobility, VPN and proxy > architectures I can think of, except for one: Mobile IP with triangular > routing. > > If you use triangular routing it might make sense to use the > care-of-address instead of the home address. But this may depend on many > things, e.g., are you mostly sending data or mostly receiving data? > So it may also be wise not to do that ... > > > > > May I ask you and the other folks here three questions: > > > 1.) Is Mobile IP with triangular routing a relevant use case? > (Or is there another relevant mobility mechanism for which Michael's > proposal does not make sense?) > > 2.) Are we really sure that it is a good idea to optimize based > on the CoA (or deviate from Michael's proposal in some other way)? > E.g., are there simulation results or field trials showing > that much better results could be achived? > > 3.) Do we really need to standardize this, in order to guarantee > interoperability (we could leave the decision up to the > implementors, as long as we are sure that letting them choose does > not harm interoperability)? > > > I think adding lots of text about mobility mechanisms in general and the > triangular routing use case in particular makes only sense if a > significant number of people answers three times "yes". > > Personally, I would answer three times with "no" ... > > > > Thanks > Sebastian > _______________________________________________ > alto mailing list > [email protected] > https://www.ietf.org/mailman/listinfo/alto _______________________________________________ alto mailing list [email protected] https://www.ietf.org/mailman/listinfo/alto
