louden@MITRE-GATEWAY.ARPA.UUCP (06/11/86)
I have a few questions about RFC 986. First, I have a LAN that is both connected to a PDN and the ARPANET. Which addressing method do I use? Second, what does TCP do for a pseudo header (It contains the IP addresses) when there are no IP address in X.121 across a PDN? This may be a can of worms better addressed by saying you use all ISO or all DoD but not both mixed.
hwb@GW.UMICH.EDU.UUCP (06/12/86)
Let me make a few comments about some of the reasons behind RFC986. The question of mixing or not mixing protocols is an important one, though not really, or at least only to some extend, the issue here. What needs to be accomplished is that existing network backbones can be utilized for ISO protocols, which includes the ISO CLNS. It probably does not make too much sense to have, e.g., an Arpanet and an ISO-Arpanet existing and being disjoint at the same time. Not even on an iterim base. It therefore becomes necessary to utilize existing backbones (good examples here are the Arpanet and the upcoming NSFnet) for both protocol suites, especially since these suites do not differ that much from each other anyway. RFC985 points out quite nicely that Ethernets are some kind of lowest common denominator to which most hosts can connect. We can probably safely assume that most gateways drop their traffic onto a local Ethernet. If, e.g., two gateways, which are attached to the Arpanet would talk both IP and the ISO-CLNS, while eventually utilizing the same routing data base local to the gateway, people could talk the ISO-CLNS accross the country. It would not matter at all what runs on top of ISOgrams in this case. Further routing through subsequent gateways within the local net would obviously be a local issue here. These (sub)gateways would have to understand both 'gram protocols, too, and, e.g., could distinguish between them according to different Ethernet type fields seen on the Ethernet. RFC986 is a draft only and probably needs further refinements so that and Internet standard could emerge which could be given to the implementors of gateways. Suggestions for these refinements would be welcome. A small committee chaired by Phil Gross (Mitre) was furthermore tasked with coming up with suggestions for possible scenarios for RFC986, which will (hopefully) result in a subsequent RFC. Input for this would be appreciated, too. -- Hans-Werner -------
louden@MITRE-GATEWAY.ARPA.UUCP (06/12/86)
If the goal is to allow both ISO and DoD on the ARPANET, why not use the same link level and different net levels? This would be compatable with both protocols since X.25 is in both. Connecting the two worlds can then be done at the applications level by application gateways that understand what is needed, such as X.400 to SMTP. The issue of gateways can be handled by implementing both DoD and ISO net level protocols and selecting the one that makes sense for a packet, such as checking version number is correct. This method works in a well defined way (all the protocols are spec.ed) and does not require the large development effort to be wasted on a tempory patch until a complete transition is made. Note that changing to ISO net layer requires all hosts to modify their software even if they will be replaced before the network converts fully to ISO. Dual mode operation would only require gateways connection dual mode networks to be modified. This sounds cheeper to me. To my knowledge there are at least 3 organizations working on this solution. The largest and best funded is the NBS effort to develop a standard application gateway. This has been discussed at the NBS ISO workshops. ... Sorry I cannot reply directly to you because we are using a BBN C70 with the standard release software. This does not support domain addresses other than .ARPA. BBN says they will fix as soon as they can but it is an example of the problems in requiring a conversion in existing protocols. It sometimes takes more time and money than you have.
hwb@GW.UMICH.EDU (06/12/86)
I think we are really talking about the same thing (utilizing the same link layer). Please re-read my previous message. The issue here is *routing* and not so much how protocols get used. The need for application gateways for the interim is definitely implied in here. -- Hans-Werner PS: NBS and ISO folks (among others) reviewed the RFC986 before it got published. -------
louden@MITRE-GATEWAY.ARPA (T. Michael Louden, MS W422) (06/12/86)
So the RFC is for addressing in an ISO stack on the ARPA style net. That makes more sense to me. Thanks for the clearification!