Hi Robert, On Mar 30, 2012, at 6:07 AM, Robert Raszuk wrote:
> Hi Darrel, > > My point was not to say that DDT is a wrong approach. I am just looking from > user POV and wanted to see why DNS infra (which is relatively easy to reuse) > would not offer me the off the shelf open source tools to build a mapping > plane. Do you really expect that every SP potentially interested in LISP will > develop it's own mapping plane code ? No of course not, I'm sorry if I gave that impression. I expect (hope?) there will be implementations from a variety vendors, as well as open source implementations. > > If you look at bigger picture we are right there in the IETF with the need > for inter-as globally scoped information distribution. Some push to use BGP > new SAFI as an overlay for CDNI, some put more data onto existing DNS, yet > some develop new point protocols (DDT). That clearly proves to me the need to > have a common service bus .. perhaps reusing a lot of BGP attributes however > allowing very easy flexibility for new types of information transport without > need to go to IETF and each time argue for 2 years to justify new SAFI. Personally, I like the fact that there are a wide set of solutions to a variety of different problems. But I can see how others might find the current environment an opportunity for consolidation. > > Openlisp as far as I see have no mapping plane code available. I have heard from people here in Paris that they are on open source DDT Mapping system code, and I was told that they welcome contributions. :-) > I guess the ALT failed as it would have to include full BGP. I don't think the ALT failed per say (though, admittedly, I'm biased :-) ). But I do think it was found to have some properties that operators found costly and we wanted to do better. As I said in my earlier message, a part of that cost was the use of a protocol that wasn't explicitly designed for LISP. The original charter of the LISP WG explicitly encouraged further mapping system experimentation, something the authors of LISP+ALT have been enthusiastically working on. > Maybe with DDT you will have more luck. Contrary openlisp.org has ITR code > which practically I would rather have hardware based. Mapping plane is x86/vm > appliance based so it would be much more important to download and run (if > one needs LISP). > I agree that it makes sense to have solutions in the mapping plane that can be independent of hardware forwarding implementations. > Last I was not clear from any presentations how DDT would inter-operate with > ALT. Well perhaps there is no issue with migration yet however as we are at > the experimental mode and it would be great to see how easy/difficult is to > migrate one mapping plane to the other. We spent some time talking about interoperability and decided that it was less costly to migrate directly to DDT. We migrated the LISP Beta Network (www.lisp4.net) from LISP+ALT to LISP-DDT a couple of weeks ago. The migration went smoothly, and we've gotten a good bit of feedback on the protocol that the authors hope to reflect in future versions of the draft. One thing to note is that the sites did not notice the change, since xTRs interface to the mapping system with Map-Resolvers and Map-Servers, and those interfaces did not change. We have recently added a second, independent to the Beta Network, lisp service provider network to the LISP-DDT mapping system, and have plans for connecting other LISP networks in the coming weeks. For more information about the ongoing deployment of LISP-DDT, you can check out ddt-root.org. Regards, -Darrel _______________________________________________ lisp mailing list [email protected] https://www.ietf.org/mailman/listinfo/lisp
