csg@pyramid.pyramid.com (Carl S. Gutekunst) (11/08/90)
>I was back in the X.25 world this morning and was reminded (incidentally) >that various PDNs have interesting requirements for X.25 implementations. >I was wondering (for the sake of curiosity) how one finds out about these >differing requirements, apart from the obvious methods of asking each PDN >and/or packet switch manufacturer. What you are really getting into here is network conformance testing. This is the hot area of X.25 (the standard itself having become stable), and there is much evolution going on; so anything I say now will likely be different in a day or two. Some terminology: CERTIFICATION is the process whereby you apply to a PDN for connection to their network. A CONFORMANCE TEST is a procedure for testing a particular X.25 implementation to determine whether it responds correctly to a variety of stimuli. Conformance tests can be standard or proprietary. Each PDN has their own unique requirements for certification. Some PDNs (like Austpac) have few requirements. American PDNs generally require a conformance test to be run on your software; European and Japanese PDNs usually require a large quantity of paperwork plus compliance tests of hardware and software. Understandably, the paperwork usually must be written in the country's native language. In countries with privatized or semi-privatized phone systems, like Japan and the U.K., hardware and software are regulated by different (possibly competing) organizations. Originally, each PDN had its own proprietary conformance test, the details of which were usually a closely guarded secret. You might get a vague specifica- tion document, but that was hardly enough to determine what oddities existed in the network. In fact, several of the PDNs I contacted were indignant that I would imply that such a document was necessary. From one I inferred that they were trying to discourage vendors from developing interfaces that passed the tests but were not truly X.25 compliant. (If you ever looked at the X.25 software out there, you'll understand that they had good cause for this.) Most PDN conformance tests were run on a live network. Telenet (er, SprintNet) was one of the first to use a test run on-site using a protocol analyzer; you purchased or rented the equipment and a box of tapes full software, ran the test, and sent the results in by mail. The Telenet test was also wonderful in that it told you all the details of the test. Good thing, too, since to pass Telenet you have to do things that would fail any other X.25 conformance test. Most European PDNs are less trusting: they have replaced online testing with protocol analyzer testing, but you have to pay for their people to come to your site to run the test. Accunet has a nice compromise: they run the test using an analyzer, but through a synchronous dialup modem. The beauty of this is that it costs you nothing at all; not even rental fees for test equipment. The Accunet test is also more thorough than usual. As X.25 has stabilized, a number of standard conformance tests have come into use. One of the oldest is Federal Information Processing Standard Publication 122 (FIPS PUB 122). This test was drafted by NIST (then NBS) to test for conformance for the U.S. Government's X.25 specification, FIPS PUB 100. The U.S. Military drafted their own conformance test, based on FIPS PUB 122: DCA 370-P195-5, "DDN X.25 Qualification Procedures." FIPS PUB 122 was based on X.25 1980; both it and the DDN test rigidly enforce details of X.25 1980 and will flunk a proper 1984 or 1988 implementation. The FIPS test also suffers from bugs, and both suffer from a failure to ensure that the device being tested stays in the correct test state. (For instance, FIPS generally ignores the effects of T1, T3, and N2.) Overall, the DDN test is superior, although both are seriously wanting. Two international standard conformace tests are now becoming available. As part of the process of merging the European economies, the Conference of European Posts and Telecommunications Administrations (CEPT) drafted a series of conformance tests, called "Normes Europeennes de Telecommication," or NETs. Their X.25 conformance test is NET2, and it is ostensibly based on X.25 1984. Currently at least Deutsche Bundepost accepts NET2 as its compliance test, and I understand that many other PTTs are or will be accepting it as well. You still have to pay the PTT to come out and run the test, but at least you have something you can run on your own first to ensure that it runs error free. The problem with NET2 is that it is far and away the weakest conformance test I've ever seen. With two exceptions (handling of P bit on SABM, and the MLP protocol) it is completely ignorant of the differences between X.25 1980 and 1984, or even 1976. Unlike FIPS and DDN, it never exposes the implementation under test to blatantly illegal conditions; these are usually where equipment fails, not on the regular clean-as-a-whistle packets. The *real* standard conformance test is ISO8882. This is composed of three documents, designated ISO8882-1 (Overview), ISO8882-2 (Datalink Layer), and ISO8882-3 (Packet Layer). This is a massive work, over 1000 pages, but fairly straightforward to read and implement, and *very* thorough. It also contains rules for X.25 1980, X.25 1984, ISO7776, and ISO8882. Clearly some ideas were borrowed from FIPS 122, but the end result is quite different. Rumor has the NIST and SprintNet will soon adopt ISO8882 to replace FIPS PUB 122 and their proprietary tests, respectively. I certainly hope they do, thus eliminating a few more redundant "standards" from the world. I also hope that the European PTTs will realize just what a crock NET2 is, and follow ISO8882. There are still other tests: CTS/WAN and XCTS (Corporation for Open Systems), to name two. I'm not familiar enough with them to comment much. I do know that some European PDNs use CTS/WAN as their conformance test. As far as actually running the tests.... weeeeell, that can be a problem. The DDN test is available for the Tekelek Chameleon II and 32, and the IDACOM PT300 and PT500; but DDN and BT Tymnet only accept the version on the Tekelek. FIPS 122 is available on the IDACOM, the Atlantic Research Commstate II, and on a little box from Hard Technology; the GSA only accepts the version on the IDACOM, which is a good thing since the other two versions are garbage. NET2 is available on the IDACOM and the Hewlett-Packard 4954; each is accepted by different bodies, although the IDACOM seems to be more widely recognized. ISO 8882 is advertised in IDACOM's literature, but they won't give it to you yet. Atlantic Research has it running on their Commstate 7700 Turbo, in beta test. Note that IDACOM is now part of HP, and Hard was bought out by Tekelek. Allow me to get on my soapbox: I've run compliance tests for darn near every- thing, on darn near everyone's analyzer. And the ISO8882 test on the AR Comm- state 7700 is the first I've ever used that actually behaves like production software. *Everything* else was a monumental pain in the petunias; the AR was a real pleasure. We're talking pitch black midnight in the swamp versus bright summer sunshine in the meadow. I'd really like to see NIST and the Commerce Department select the AR 7700 as their reference machine for ISO8882. <csg>