Those of us who have been doing this for a while sometimes forget that there are things that are not obvious to those just joining us. Protocol number assignments are at http://www.iana.org/numbers.html. Application procedures are at www.iana.org. IP Protocol numbers are at http://www.iana.org/assignments/protocol-numbers To answer the first question - for proprietary protocols, don't apply for your own number, use one of the already registered "any *" protocols. E.g. 9 - IGP - Any private interior gateway protocol, 61 - any host internal protocol. You should build sufficient robustness into your protocol to ignore other private protocols of the same type - e.g. a magic number in the payload, or Joe's recommendation of a private checksum algorithm. Topic closed.... Mike At 10:27 PM 3/4/2002 -0800, Joe Touch wrote: >Vernon Schryver wrote: >>>To: marelines@yahoo.com, ietf@ietf.org >>>Cc: nabe@iij.ad.jp >>>Subject: Re: Proprietary IP Protocol Type >>>From: Kuniaki Kondo <kuniaki@iij.ad.jp> >> >>>>I am working on a distributed router and i want to run my own >>>>proprietary protocol inside over >>>>the IP layer. ... >> >>>I also have a same question. >>>I would like to get a document which is described assignment criteria, >>>current being used, etc... >>> >>>If someone know, please advice me. >>Please do not be offended, but if you do not already know the answers >>to such questions or at least better places than the main IETF mailing >>list to ask them, then it is practically certain that you would be >>better served to use UDP/IP instead of raw IP. > > >>The number of available IP protocol numbers is infinitesimal compared >>to the number of UDP port numbers. (The host addresses can be seen >>as qualifying port numbers, but protocol numbers are global.) > >For prototyping, you don't need to worry about port or even protocol >number collisions. Use a checksum to ensure that you're interpretation of >incoming packets is correct, and ignore those whose checksums err. Unless >the protocol number is in common use, the overhead will be sufficiently low. > >Joe