FILLMORE@EMRCAN.BITNET (10/18/89)
I have a (perhaps naive) question about the interoperability of routers: We want to join Ethernet LANs in two different cities with a 56kbps sync line. The LANs run a variety of protocols such as XNS, TCP/IP, Decnet. We want to use routers at each end of the line. Do the two routers have to be from the same manufacturer? If so, why? One of the routers is a Cisco box but the other could be Proteon or some other make (or does it have to be a Cisco?). ________________________ Bob Fillmore, Systems Software & Communications BITNET: FILLMORE@EMRCAN Computer Services Centre, Voice: (613) 992-2832 Energy, Mines, & Resources Canada BIX: bfillmore 588 Booth St., Ottawa, Ontario, Canada K1A 0E4
MAP@LCS.MIT.EDU (Michael A. Patton) (10/19/89)
Date: Tue, 17 Oct 89 17:42:00 EDT From: <FILLMORE%EMRCAN.bitnet@ugw.utcs.utoronto.ca> I have a (perhaps naive) question about the interoperability of routers: This question is not really naive, in fact there is a real deep understanding of "compatability" involved. We want to join Ethernet LANs in two different cities with a 56kbps sync line. The LANs run a variety of protocols such as XNS, TCP/IP, Decnet. We want to use routers at each end of the line. Do the two routers have to be from the same manufacturer? If so, why? One of the routers is a Cisco box but the other could be Proteon or some other make (or does it have to be a Cisco?). The simple answer is that TODAY the two gateways at opposite ends of a serial line have to be from the same company. Presently the framing used on serial lines by the various gateways is different. There are a few standards for this, but they aren't comprehensive enough. Each vendor had some feature or facility they needed to support which did not fit and they each developed a framing system independantly. This means the framing is not compatible. There is presently a Working Group within the IETF (the Point-to-Point Protocol [PPP] WG) which is addressing this problem and it should be issuing an RFC RSN. Drew Perkins gave a presentation on the protocol at InterOp '89. When this gets finalized all (most?) of the gateway vendors will be implementing it so that you can mix them as you describe, but it won't be possible before then. __ /| /| /| \ Michael A. Patton, Network Manager / | / | /_|__/ Laboratory for Computer Science / |/ |/ |atton Massachusetts Institute of Technology Disclaimer: The opinions expressed above are a figment of the phosphor on your screen and do not represent the views of MIT, LCS, or MAP. :-)
cire@CISCO.COM (cire|eric) (10/20/89)
Bob, This is not at all a naive question. Both routers have to currently be from the same manufacturer. That is because they are on the end of a serial line and there currently isn't an agreed upon standard for how to talk across these lines in general. You could use X.25 and in some cases that would work for different vendor's equipment but that is shall we say a less than optimal solution. There is currently an effort under IETF auspices to standardize a decent way of communicating and interoperating across serial lines. It is called the Point-to-Point Protocol (PPP) and should see light of day sometime later this year or in 1990. Cisco plans on supporting it and from what I understand others will as well. -c cisco engineering
medin@NSIPO.NASA.GOV ("Milo S. Medin" [NASA ARC NSI Project Office]) (10/21/89)
That's not exactly true. I believe the routers will interoperate at least for IP if you use X.25 as your serial line protocol. Both Proteon and Cisco support X.25 framing now. It's not a very nice solution, but it should work. Both Proteon and Cisco have committed to implement PPP, and both companies had active participation in the working group. Thanks, Milo
CERF@A.ISI.EDU (10/22/89)
Today it is a likely that you would need routers from the same vendor, linked by pt-pt channel to provide you the multiprotocol functionality you want. The IAB and IETF are working towwards the development of an Internet standard for multiprotocol routing which could be supported by all vendors of routers. Vint Cerf
MAP@LCS.MIT.EDU (Michael A. Patton) (10/22/89)
Date: Fri, 20 Oct 89 11:28:07 -0700 From: "Milo S. Medin [NASA ARC NSI Project Office]" <medin@nsipo.nasa.gov> That's not exactly true. I believe the routers will interoperate at least for IP if you use X.25 as your serial line protocol. Both Proteon and Cisco support X.25 framing now. It's not a very nice solution, but it should work. That's not my understanding. They both use X.25 framing to set off the packets on the line, but they use different encapsulation to provide header compression, protocol identification, etc. This is sort of analogous to ISO 802.3 vs Ethernet V2. They are electrically compatable and the low level framing is the same, but if one machine speaks 802.3 and another speaks only Ethernet V2, they won't communicate. Packets sent by one are rejected as mal-formed by the other. Both Proteon and Cisco have committed to implement PPP, and both companies had active participation in the working group. That's correct. Basically the design of PPP is intended to give enough flexibility to let them have any special handling they need and a compatable framework to allow any two machines to work together with whatever common features they implement. Mike Patton
ESC1814@ESOC.BITNET (dave stafford) (10/23/89)
Hi Milo, Why isn't X.25 a nice solution? We use it between sites here and it makes for a convenient solution, not only for routing but to connect different flavours of routing devices. Dave Stafford Networking Eng. European Space Operations Centre Darmstadt, W. Germany
finco@ficc.uu.net (gary finco) (10/23/89)
In article <89Oct17.165048edt.57354@ugw.utcs.utoronto.ca>, FILLMORE@EMRCAN.BITNET writes: > I have a (perhaps naive) question about the interoperability of routers: > We want to join Ethernet LANs in two different cities with a 56kbps sync > line. The LANs run a variety of protocols such as XNS, TCP/IP, Decnet. > We want to use routers at each end of the line. Do the two routers > have to be from the same manufacturer? If so, why? One of the routers > is a Cisco box but the other could be Proteon or some other make > (or does it have to be a Cisco?). > ________________________ After talking to both Cisco and Proteon about the same question, I was told that it is recommended that the routers be from the same manufacturer at both ends of the link. The reason they gave me was that they have their own data link implementation which most likely will not match exactly with a different vendor's. I also feel that having one vendor per link will eliminate the finger pointing that could arise from a failure to communicate. Disclaimer: I speak for myself and no one speaks for me. gjf
henry@utzoo.uucp (Henry Spencer) (10/24/89)
In article <8910181843.AA23496@gaak.LCS.MIT.EDU> MAP@LCS.MIT.EDU (Michael A. Patton) writes: >There is presently a Working Group within the IETF (the Point-to-Point >Protocol [PPP] WG) which is addressing this problem and it should be >issuing an RFC RSN... Is there a current estimate of the first digit of the calendar year in which the PPP RFC will appear? It's been "presently a Working Group" and "addressing this problem" and "RSN" for a longish time... -- A bit of tolerance is worth a | Henry Spencer at U of Toronto Zoology megabyte of flaming. | uunet!attcan!utzoo!henry henry@zoo.toronto.edu
almes@RICE.EDU (Guy Almes) (10/24/89)
It is well known that you can't currently put different makes of router on different ends of a serial line. The easiest explanation is data link and encapsulation. Will support by the various vendors of the coming point-to-point protocol correct this situation? I know that it's intended to be a move in the right direction, but is there a claim that it will be sufficient? I'd like to hear from people on the PP protocol IETF WG, and from knowledgeable others. One benchmark question: will the situation be as good as when routers of multiple makes are placed on an Ethernet? -- Guy
Gene.Hastings@BOOLE.ECE.CMU.EDU (10/25/89)
Bear in mind a potential gotcha: Once the router vendors support PPP, there is still the potential for disagreement about the encapsulation of other protocols (such as D*CN*T), and some of us need to support multiprotocol routers. Would any of the router vendors like to respond? Gene
amanda@intercon.com (Amanda Walker) (10/25/89)
In article <1989.10.24.16.57.48.Gene.Hastings@boole>, Gene.Hastings@BOOLE.ECE.CMU.EDU writes: > Bear in mind a potential gotcha: Once the router vendors support PPP, there > is still the potential for disagreement about the encapsulation of other > protocols (such as D*CN*T), and some of us need to support multiprotocol > routers. > > Would any of the router vendors like to respond? > > Gene Well, I'm not a router vendor, but I have looked over the draft PPP, and it's got a 16-bit "type" field that is basically the equivalent of the Ethernet type field (same numbers and everything, by some strange coincidence :-)). One of the features of PPP as opposed to SLIP is this very capability to run multiple protocols over a PPP link layer. Unless I have misread the draft RFC, it doesn't leave much room for disagreement... -- Amanda Walker <amanda@intercon.com>
awaldfog@wellfleet.com (Asher Waldfogel) (10/25/89)
>Bear in mind a potential gotcha: Once the router vendors support PPP, there >is still the potential for disagreement about the encapsulation of other >protocols (such as D*CN*T), and some of us need to support multiprotocol >routers. > >Would any of the router vendors like to respond? > >Gene As a member of the PPP working group and a multiprotocol router vendor, I can vouch that one of the goals of PPP is multiprotocol interoperability. We have even discussed bridge interoperability. So far only IP-oriented vendors have participated in PPP. I wouldn't count on Decnet- or bridge- only systems to implement PPP. Asher Waldfogel Wellfleet Communications
ccruss@ucdavis.ucdavis.edu (Russ Hobby) (10/28/89)
Well here is the word from the Point-to-Point Protocol (PPP)
workgroup. We will have a final draft at the end of IETF next week.
The major router vendors (as well as workstation and terminal server
vendors) have been participating in the workgroup all along. Most of
the router vendors say that they will have PPP in their next release.
PPP has been designed to handle multiple protocols, not just IP, so
DECNET and other protocols will be transported in a standard method
over the link. PPP has many other features such as link establishment,
error detection, link testing, authentication, encryption and IP
address negotiation and is extendible for adding new protocols and
features and as such there is still work to be done on the protocol.
For example only a simple user/password authentication is currently
defined. A more secure method would be desirable. Also, although the
method of encryption can be negotiated, no methods have been define
at this time.
In answer to Guy's question, it will be equilvalent to pluging the two
routers into an ethernet. PPP does not solve all the router
interoperability problems though. It does provide a standard method of
getting the packets across the wire, but it doesn't solve the routing
problem. Currently RIP can be used, but it has a lot of limitations. It
seems that OSPF will be the next improvement and it appears the
Internet community has agreed that it will be the next step. But even
OSPF will not solve all the long term routing problems. That is still
a subject for research.
There have been to independent implementations of PPP. One by
UC Davis for PCs based on the KA9Q code and the other by CMU for
4.3BSD. These implementations were done to test the protocol and to
find any weakness. As we know one implementation may be perfectly
compatible with itself but not with others, so that's why we did two
independent ones. Both of these systems were demonstrated
interoperating at INTEROP. The implementations will soon be publicly
available and we encourage porting to other systems.
Russ
Russell Hobby
Data Communications Manager
U. C. Davis
Computing Services INTERNET: rdhobby@ucdavis.edu
Davis Ca 95616 BITNET: RDHOBBY@UCDAVIS
(916) 752-0236 UUCP: ...!ucbvax!ucdavis!rdhobby