POSTEL@USC-ISIF.ARPA (05/07/84)
Robert Morris: Have your read the RFCs on domains (881, 882, 883, 879)? Other top level domains will be added. There are several requirements that have to be met for something to be a domain. First, since domains are administrative entities they have to have a responsible management - there has to be an individual in charge of the domain. Second, there have to be lookup servers for the domain data provided - this has to be very robust. Third, the domain must have some minimum size - at the top level probably 100s of hosts. Fourth, the top level domains have to be registered with the NIC. Another RFC on "domain requirements" is in the works. --jon. -------
morris@harvard.ARPA (Robert Morris) (05/08/84)
Thank you, Jon, for your letter and information. I read rfc883 and admit to being confused. How much of 883 deals with looking up host numbers (possibly to send mail), and how much deals strictly with mail? For instance, does one really need a name server for a domain which is not on the internet, and thus only accessible through a mail (only) gateway. In particular, I would like to register a top level domain with a couple of hundred hosts. There are a few people here who already have big databases describing it, and are trying to keep them up to date. We can control the format and so forth of messages going onto the arpanet. What I need is clarification of what the name servers are for; I'm not sure what in rfc883 applies to domains with a single gateway and no internet host tables. Thank you, Robert Morris (morris@harv-10)
POSTEL@USC-ISIF.ARPA (05/08/84)
Robert: Enclosed is a note i recently sent to the Namedroppers list that addresses the point you raise. (To join Namedroppers send a request to "Namedroppers-Request@SRI-NIC.ARPA".) In general, what ever information you want hosts in the ARPA-Internet to be able to resolve has to be available in lookup name servers accessible via ARPA-Internet protocols. If it is sufficient that the only thing that can be done is find the Internet Address of the relay host into your domain then the database is simply the entry mapping "*.DOMAIN" into "Relay-Host-Address". If you want ARPA-Internet hosts to be able to determine before sending mail that the destination host actually exists, then a full database to the host level is needed. If you want ARPA-Internet hosts to be able to check if the recipient mailbox is valid and to determine if there is a forwarding address etc, then a database to that level of detail must be available. --jon. Date: 3 May 1984 20:07:38 PDT From: POSTEL@USC-ISIF Subject: Inter-Enviromnent Name Service Hi: There are two parts to the domain name system. The first is the introduction of domain style names. The second is the introduction of domain name lookup service. In both cases, the design is intended to be widely applicable in a variety of communication environments, not just the ARPA-Internet. We have a reasonable expectation that the domain style names will be used in a variety of environments. We have (so far) little reason to expect that the domain name lookup service will be implemented in any environment other than the ARPA-Internet. However, for a host in the ARPA-Internet to make use of a domain style name (from any environment) that host must be able to lookup that name using the domain name service via ARPA-Internet protocols. This means that every domain style name from any environment that is to be meaningfull to ARPA-Internet hosts must be listed in some domain name lookup server in the ARPA-Internet. Suppose there were some domain (let us call it XYZ) in some environment (let us call it PQR) not even sharing any common element with the ARPA-Internet or any of the domains overlapping the ARPA-Internet, yet communication between hosts in XYZ and hosts in the ARPA-Internet is possible via some third parties. For this communication to be possible, some domain name lookup servers in the ARPA-Internet would have to be able to answer queries about host names in the XYZ domain. At first blush, this would seem to require that a complete detailed and up-to-date copy of the database of hosts from the XYZ domain would have to be maintained in a domain name lookup server in the ARPA-Internet, at locations possibly far removed from any part of the XYZ domain. But, what is the necessary information in this database? If, as is likely, all the communication between ARPA-Internet hosts and hosts in the XYZ domain is routed via a particular relay host, then all the entries in the database will point to that relay host. If it is desired to verify that a particular host name in the XYZ domain is valid, then the full database is required. If it is sufficient to find the address of the relay to the XYZ domain, then the database can be a single entry for the name "*.XYZ" with the address of the relay host. That is, any query with a domain style name ending in ".XYZ" will match the entry, and will receive a reply indicating the relay host. Please notice that the situation is symmetric. If the XYZ domain hosts used a procedure similar to that of the ARPA-Internet hosts in resolving host names then the domains overlapping the ARPA-Internet would have to provide databases describing their domains in a form suitable to the name servers of the PQR environment. --jon. -------
kevin@harvard.ARPA (Kevin Crowston) (07/22/85)
> An example of what I'd like to see: > if I want to get mail here (say, at hpfclo.FC.HP.COM), > someone should be able to send to hpfclo.COM from anywhere. Unfortunately, there may be more than one hpfclo's in the world (well, maybe not hpfclo, but perhaps VAXA? One at HP, one at MIT, one ...). Part of the usefulness of domains is that the specific name of the host need not be unique, only the full name. Furthermore, your scheme still requires someone to know about everybody on the net or that everyone on the net take a look at your message, as you route it around to increasingly smarter hosts. Also, what happens if I mail to hpfcl0.HP.COM? Since no host knows about it (it probably doesn't exist), who eventually bounces it back if everyone just forwards it to someone smarter? Giving the full name hpfclo.FC.HP.COM only requires that you can find the COM name server, which in turn can find the HP server, etc. -- Kevin Crowston UUCP: {seismo,ut-sally}!harvard!kevin MIT Sloan School of Management ARPA: kevin@harvard.ARPA
Rudy.Nedved@cmu-cs-a.ARPA (10/16/85)
First off, the domain system is a step in the right direction as compared to the centralized host table mechanism. Just like in the telephone system, you call the area code you are interested in, specify the city and then query about a name for a telephone number. It would be unmanageable and extremely slow to have a centralized telephone directory. Second, the point about domains is an experiment is misleading. The subtle issue here is that the research side of the ARPA Internet known as ARPANET has somewhat formally adopted domains. This is almost a given. Life does not contain absolutes so people can say otherwise. The production side of the ARPA Internet known as MILNET has formally indicated in one of the implementation notes a "wait and see" with no commitment. In other words, while the ARPANET is fighting over domains...they can continuing doing work....which is a good for a production enviroment...they can not fight with the problems...they have other problems. Third, the hosts in the domain system but not in the old host table are not incorrect or illegal as far as any specification is concerned. The real question is one of practicality. If a non-domain system host MUST talk to a host not listed in the domain system, then the postmasters involved or liaisons should communicate and find a solution. If these types of comprimises don't exist then the ARPANET can not experiment and the MILNET can not get work done. Given I live and maintain a very large and rapidly expanding computer system enviroment, I have to deal with both of these issues every day. At the moment, CS is experimenting and developing solutions to problems and creating reliable software. If other departments want to add people or systems to our experiements...great. When we feel things are at a production level then we expand man power and other resources to get the system installed all over. This seems to be the same thing the ARPA Internet is doing in a larger scale... -Rudy