jmn@ariel.UUCP (J.NERVIK) (10/02/84)
[] In looking at the documentation on IP, there appears to be two dialects of IP. This looks like chaos to me, and I wonder if someone with perspective on these developments would clarify the situation. One IP is specified in DARPA documents, mainly; -RFC-791, "Internet Protocol", September 1981. Critical parameters include source-address and destination-address as fixed length fields of 4 octets each. The maximum header size is 60 octets, but typically is 20 octets. All hosts must be prepared to accept datagrams of up to 576 octets. This allows for a 512 octet data field, a header, plus some octets for the higher level protocols. The second IP is specified in ISO documents; - TC 97/SC6 N, "Information Processing Systems - Data Communication Protocol for providing the Connectionless-Mode Network Service", May 1984. My copy has DIS8473 handwritten on it. - Draft Proposal ISO/DP 8348/DAD2, "Information Processing Systems - Data Communications - Addendum to the Network Service Definition Covering Network Layer Addressing", April 1984. The source-address and destination-address are variable length fields of up to 20 octets or 40 digits each. The maximum size of the header is 254 octets. Datagram size is not specified, other than to recognize that a datagram must at least hold one octet of data to have any use at all. The general features of these two IP dialects seems to be about the same. However, reading these documents is a good way to cure insomnia, and I suspect a close reading will find many differences in detail. It is my understanding that the ISO documents have been generated as part of the process to get RFC-791 accepted by the international standard bodies. The ISO documents describe a much fatter protocol. Why??? Has the IP of RFC-791 been destroyed??? If ISO adopts these documents, will ARPANET and/or MILNET switch to the ISO version of IP??? John M. Nervik
mark@cbosgd.UUCP (Mark Horton) (10/05/84)
My understanding of the situation (containing mostly information from the DARPA side of the fence) is that the ISO IP (I sure hope they change the acronym, we're already getting confused) was not intended to be compatible with the DARPA IP. Much of the functionality was copied in an attempt to get DOD to conform to the upcoming ISO standard. However, this is largely a political game, and the last I heard there was no real intent on the part of DARPA to switch over to the ISO protocols. It is possible this will change when ISO finally gets its protocols documented. My own personal opinion is that ISO will be out too late. The virtual terminal protocol (the rest of us call this remote login) won't even be in draft form until 1986, and you can't have a usable network without remote login. (I don't consider UUCP exactly a usable network either.) It could easily be 1988 or later by the time the VT protocol is approved and vendor implementations start to appear. By that time, TCP/IP will have gathered so much momentum it will be impossible to stop it. And network compatibility being what it is, there will be too many existing TCP/IP networks out there to attempt to convert them or gateway between them with any high level of functionality. I suspect that ISO will catch on primarily outside the USA, and within the USA will only catch on to the extent that X.25 has (although it will have a fight to convert the X.25 networks over, so it may not do even that well.)
jbn@wdl1.UUCP (jbn ) (10/13/84)
ISO IP and DoD IP are not the same protocol. ISO IP allows for longer
addresses, among other things; the consensus is that 32 bits is not enough for
a worldwide address space. It is supposedly possible to run TCP on top of
ISO IP. There is also an ISO Transport Protocol, which is quite different
from TCP but like TCP provides a byte-oriented virtual circuit. It is based
on the National Bureau of Standards Transport Protocol, but there are
significant differences. ISO Transport is very much a compromise, like
X.25, and there are lots of optional operating modes to placate the datagram
and virtual-circuit factions. ISO Transport was used at the big GM-sponsored
demo at the National Computer Conference last month. The present spec for
this protocol is not very tightly drawn as yet; there is talk of using a
formal technique to specify the protocol.
Incidentally, the official DoD definitions of IP and TCP are not the
RFCs, but MIL-STD-1777 (IP) and MIL-STD-1778. The MIL-STD versions are
written in an Ada-like psuedocode. A revision of MIL-STD-1778 is underway
with the objective of tightening up the spec to the point that compliance
insures interoperability. The object is not to change anything, but to
bind certain important decisions left unspecified in the present document.
In other words, if A can't talk to B, then we should be able to tell
whether A is wrong or B is wrong. If we can't, the spec is wrong.
The person formally in charge of the revision is Mike Corrigan (Corrigan
@DDN1) of the Defense Data Network program office.
John Nagle
Ford Aerospace and Communications Corp.
ps: Please do not ask me for copies of the MIL-STDs; contact NTIS.sylvain@lvbull.UUCP (Sylvain Langlois) (10/17/84)
(< Crunch, Crunch...>) Let me join the game! We , here in Europe, have got what is called a "Standardization Body" called ECMA, which is supposed to gather all the european Computers Manufacturers (By the way, IBM is sitting at the table....!). This Organisation has written a document called ECMA 92, which is at this moment a working document (still DP or something equivalent!) and which talks about an ... Internet Protocol. And guess what... IP seems to be a subset of this Protocol!! I don't know exactly what's going on in the ISO house about this subject, but I've read somewhere that the ISO "IP" should be voted to DIS in the very near future!! So there should be a document about it somewhere!! The thing is that ECMA and ISO are working hand in hand and reading ECMA 92 give me a good idea of what should be ISO Internet Protocol. Of course, I know the ISO side better than the DARPA one, but I promise I will IP specs better... An other point is that NATO (hope I spelled it right, we call it OTAN over here!) has decided to switch to ISO Architecture in the near future. I guess it will influence ARPA too!! I'm interesting in what is going on at DARPA now, could anybody (Mark?) give me more infos. Thanx in advance... ---- Sylvain Langlois (...mcvax!vmucnam!lvbull!sylvain) --- PS: We are having problems with incoming news sometimes, mail answers are welcome (as a backup of news articles!).