kiessig@fortune.UUCP (11/03/83)
Has anyone out there heard of NS or Courier implementations for UNIX? Rick
skip@gatech.UUCP (11/06/83)
The CCITT standard X.25 is not for use within a network, but as an interface
between a packet switched network and devices wishing to attach to it, such
as host, terminal concentrators, etc.
The TCP/IP protocol (which I know the least about, so if I make a mistake,
please forgive me) is formulated by the National Bureau of Standards and is
blessed by the Defense Department. I'm not sure if it covers all the layers
of the ISO model. Because of the DoD blessings, many companies are at least
implementing gateways to TCP/IP.
The Xerox Network Systems protocol (XNS) is being adopted by a plethora of
companies who aim to be compatible with each other on an Ethernet-based
cable and access scheme. The XNS protocols include as part of them the
Courier protocol which covers two (I think -- I've got the documents back
at the office) of the seven layers. I think it covers the Network and
Transport layers (3 & 4). All layers have been published except for the
necessary routines to access the Xerox laser printers.
Of course, there are many other proprietary protocols listed which serve
various purposes...
-- skip addison {akgua,allegra,rlgvax,emory}!gatech!skipmark@cbosgd.UUCP (Mark Horton) (11/07/83)
Skip, if you have a published copy of the higher level layers for XNS (e.g. everything above the transport level) I'd sure like to hear about it. My understanding is that Xerox will not release the specs for the higher layers, because they want you to buy Xerox computers that implement the layers and not to be able to use any other vendors computers. This alone is the kiss of death, as far as I'm concerned. Of course, if I'm wrong I'd like to know about it. My Xerox literature claims they have a 5 level architecture, with only levels 0, 1, and 2 (physical, internet datagram, and 5 protocols including sequenced packet and packet exchange) published. Level 3 includes courier (which I gather is a session layer protocol) and implies that all the application layer protocols do not matter. (I see this attitude with a lot of vendors - they think you are going to provide your own application level software and make not attempt to standardize it.) My impression is that the only network protocol suite that documents the standards all the way up to the application layer is TCP/IP. They have a standard for remote login; a standard for file transfer; a standard for electronic mail; a standard for a nameserver; even standard for Mickey Mouse protocols like time of day, who is logged on, a quote of the day, data sink, data generator, echo server, and so on. This does not provent you from inventing new application protocols (e.g. Berkeley UNIX has added remote execution) but standardizes the ones that everybody is going to want. I'm sure they must exist with XNS too, but they do not seem to be publicly documented. Mark Horton
skip@gatech.UUCP (11/08/83)
While the Internet transport and Courier protocols don't define the application
layer, I thought Xerox had released the information for all the applications
except the print service. In any case, Xerox has contracted with other
companies who wish to develop software using Xerox's application layer
protocols. For example, I expect within a few months to see DEC announcing
support of the Interpress (print service) protocol so that your Vax hooked
up to an Ethernet could make use of a Xerox print server. I've only heard
that from one source, but it wouldn't surprise me. Xerox has got to make
money somehow. Selling printers may be the only way. (I doubt they've
made money on the Star system.)
-- skip addison {allegra,rlgvax,akgua,emory}!gatech!skipjsq@ut-sally.UUCP (John Quarterman) (11/09/83)
I was hoping somebody else would follow this up, but since nobody has: >The TCP/IP protocol (which I know the least about, so if I make a mistake, >please forgive me) is formulated by the National Bureau of Standards and is >blessed by the Defense Department. I'm not sure if it covers all the layers >of the ISO model. Because of the DoD blessings, many companies are at least >implementing gateways to TCP/IP. TCP/IP is not one protocol, but two: IP (Internet Protocol) is the top of ISO's network layer (3) and handles internet addressing, among other things; TCP (Transmission Control Protocol) fits in ISO's transport layer (4) and handles fragmentation/reassembly, process-host/process-host reliable connections, etc. There is a datagram protocol called UDP that can be used on top of IP in place of TCP. TCP/IP was developed for the Defense Advanced Research Projects Agency by numerous research institutions, such as BBN, Rand, ISI, Berkeley, etc. on the ARPANET over many years. Work continues, mostly being conducted via the ARPA Internet. The National Bureau of Standards has adopted TCP/IP in the last few years (though they've changed their version somewhat). TCP/IP and related protocols do cover all the ISO layers, though they were developed under a different reference model than the ISO one. The ISO model is a later one built partially on ARPANET experience, but with somewhat different goals and perspectives. There is a paper on this subject by M.A. Padlipsky, "A Perspective on the ARPANET Reference Model," that might be of interest. The reason for the increasing industry use of TCP/IP is not so much DoD blessings (more of a curse to many), but the plethora of existing implementations of TCP/IP. There are currently about 130 registered networks and 780 hosts in the ARPA Internet, all running TCP/IP. This may not seem impressive compared to the 1600 odd UUCP sites, but the quality and variety of service is several orders of magnitude better. The underlying physical networks include ARPANET-style (BBN 1822) networks, ethernets, and token rings. The host machines and operating systems are too numerous to mention. In other words, it's been around longer than most anything else, it has great adaptability, and it works. -- John Quarterman, CS Dept., University of Texas, Austin, Texas {ihnp4,seismo,kpno,ctvax}!ut-sally!jsq, jsq@ut-sally.{ARPA,UUCP}
mark@cbosgd.UUCP (Mark Horton) (11/10/83)
Here are two vendors that do XNS for UNIX: Fusion, from Network Research orp in LA, (213) 474-7717 Bridge Communications, Silicon Valley, (408) 446-2981 (not really for UNIX, but plug-in to various ports or multibus) I don't know if either are any good. Fusion has a few application programs (file transfer, remote login, printer, random) which probably do not conform to any particular standard. Bridge has gateways over ethernet and X.25 but seems to offer nothing higher level than transport.
sanders@menlo70.UUCP (Rex Sanders) (11/15/83)
I don't have the mfr literature at home to verify, but I believe Interlan is at least promising XNS for Unix. Also, just today I read in Byte about 3Com using Vax/Unix systems as fileservers for their IBM PC/Ethernet setup. Since the PC talks XNS, I presume they must be doing some form of XNS for Vax/Unix. This means that 3Com is supporting both TCP/IP and XNS on Vax/Unix systems. Can't they make up their minds? -- Rex ucbvax!menlo70!sanders
chuqui@cae780.UUCP (Chuq Von Rospach) (11/17/83)
Bridge Communications, Silicon Valley, (408) 446-2981 (not really for UNIX, but plug-in to various ports or multibus) I don't know if either are any good. Bridge has gateways over ethernet and X.25 but seems to offer nothing higher level than transport. When I looked rather carefully at Bridge a few months ago, they were limited to transport layer only. They were doing development on a Vax, but they couldn't even use their own ethernet to download software into their boxes for testing (their ethernet boxes use 68K machines). At that time their system looked like a good way to connect things together when you had a number of users and a number of machines that needed to connect in relatively random ways. It was possible to get one machine to connect to another machine and transfer files, but that software was up to the user. (Bridge's system, the CS/1, talked from RS232 to RS232 across an ethernet. It did not talk to (for example) an ethernet board in a vax. In this months (November) Mini-Micro systems, they are announcing their CS/100, which is a single board (and cheaper) version of the CS/1. It seems to be able to support 10 RS-232 lines, where the CS/1 supports up to 32. They also mention optional software for file transfer for major OS's (CP/M, MS-Dos, VMS, and Unix) that they didn't have before. This is probably at the application level on the machine (since connection still seems to be through RS-232), but at least its there. I was impressed with the engineering and (from what I saw) stability of their systems. My only quibble at that time was that it seemed like a rather expensive way to put together a port contender system. Now with the FTP software, it becomes much more general purpose (and with the CS/1, less expensive). -- From the dungeons of the warlock: amd70!cae780!chuqui Chuqui the Plaid *pif*
mark@cbosgd.UUCP (11/21/83)
Is there any reason to believe that Bridge's FTP will talk to anybody else's FTP (past, present, or future) on XNS? That is, did they roll their own or is there some use of a standard in there?
chuqui@cae780.UUCP (11/22/83)
I believe the Bridge rolled their own. Their main design assumption was
that they are the only devices on the net. I believe they claim that they
will not interfere with other devices sharing the net (assuming they keep
to the proper protocols) but they also don't communicate with them.
--
>From the dungeons of the warlock: amd70!cae780!chuqui
Chuqui the Plaid *pif*