RFC 1291
Mid-Level Networks Potential Technical Services
RFC 1291 🎀 Mid-Level Networks: Potential Technical Services
Okay bestie, here's the tea on this one 💅
- Author: V. Aggarwal, JvNCnet Computer Network
- Date: December 1991
- Status: it provides information for the Internet community. It does not specify an Internet standard 📚
- Obsoletes / Updates: the header doesn't list either one, so nothing to flag here 🙈
Status of This Memo ✨
This RFC is giving "just here for information" energy 📚 It provides information for the Internet community and does not specify an Internet standard. Distribution of this memo is unlimited 🦋
Abstract 🌸
So the abstract basically says this doc proposes a set of technical services that each Internet mid-level network can offer, both inside the mid-level network itself and to its peer networks. "Mid-level" is the generic term for all regional and similar networks. Because of continuous evolutions and transitions, they can no longer really be called "regional" [MAN] 🌍 It also talks through the pros and cons of offering these services, plus areas where mid-level networks can work together 🤝
A big chunk of the ideas come from discussions in the IETF Operational Statistics (OPstat), User Connectivity Problems (UCP) and Network Joint Management (NJM) working groups 👑
Table of Contents 🍓
1. Introduction.................................................. 2
2. The Generic Model............................................. 2
3. Technical Services............................................ 3
3.1 Domain Name Service......................................... 3
3.2 Public Domain Software...................................... 4
3.3 Network Time................................................ 5
3.4 Network News................................................ 5
3.5 Mailing Lists............................................... 6
4. Experimental Testbeds......................................... 6
5. Network Information Services.................................. 7
6. Network Operations............................................ 7
7. References.................................................... 8
8. Security Considerations....................................... 9
9. Author's Address.............................................. 9
Appendix A Mailing Lists......................................... 10
Appendix B DNS Architecture Strategy............................. 101. Introduction 💖
Over the past few years, the Internet has grown into a seriously massive entity, and its dependability is critical to its users. Because of the size and nature of the network, the trend has been to decentralize as many network functions as possible, like domain name-service, whois, etc. 🌸 Work is also being done in resource discovery [SHHH90] so that researchers' stuff doesn't get lost in the huge volumes of data on the Internet 🔍
A side effect of all this growth is the logical structure the Internet has picked up, where networks get classified by function. The NSFnet national backbone, the mid-level/regional networks and campus networks are all real examples right now. Each one can be seen as a hierarchy within an organization, and each does a slightly different job (campus LANs give access to local resources, mid-level networks give access to remote resources, etc.) 🧩 The functions of each hierarchy then become the "services" offered to the organizational layer below it, which in turn depends on those services. It's giving layers, honestly 🍰
This document proposes a set of basic technical services that a mid-level network could offer. These services would not only make the mid-level network itself more robust, they'd also help structure how resources and services get distributed across the Internet ✨ It also proposes a uniform naming convention for locating the hosts that offer these services 📛
2. The Generic Model 🧚
The Internet model this document is built on is a graph of mid-level networks connected to one another. Each one connects the campus/organization networks, and the end users are attached to the campus networks 🕸️ The model assumes the mid-level networks are the highest level of functional division in the Internet hierarchy described above (this could change in the unforeseen future) 🔮 With that picture in mind, this document is all about minimizing unnecessary traffic on the Internet and making the whole structure as robust as possible 💪
The proposed structure is basically an extension of organizational LANs, where certain services live inside the organizational LAN itself. Think nameservice, mail, shared files, and single or hierarchical points of contact for problems, etc. 🫶
The following are the services discussed as possible functions of a mid-level network:
- Technical services
- Experimental sites for testing and dissemination of new software and technology to end sites on the network
The following services are also mentioned briefly, since they're discussed in more detail elsewhere [SSM91, ML91]:
- Network Operation services, and the interaction between different mid-level networks in this area
- Network Information services
3. Technical Services 🎀
The Internet has become an essential entity because of the services it offers to its end users. The list is long and growing, but some services are way more widely used and deployed than others 📈 This section tries to list and discuss the technical services that could help a mid-level network provide robust, improved services to its end sites 💅
3.1 Domain Name Service 🌐
According to the NSFnet traffic statistics collected for May 1991, about 7% of the packets on the NSFnet backbone were domain nameserver (DNS) packets. That's a big amount of traffic. Since most other network applications depend on this service, a robust DNS service is critical to any Internet site 😭
Placing secondary nameservers on different physical networks can make this service a lot more reliable [MOC87a, MOC87b]. Here's the catch, though: the nameservers for the next highest level have to be available in order to resolve names on the outline-mode side of one's domain. So for foo.princeton.edu to resolve a.mid.net, the root nameservers pointing to mid.net's nameservers have to be reachable 🫠
To make things more reliable, the mid-level network could have at least one nameserver that can resolve nameserver queries for all domains directly connected to it. That way, if the entire mid-level network gets cut off from the rest of the Internet, applications can still resolve queries for sites directly connected to it 🛟 Without this, there's no way to resolve a name if the root (or higher level) nameservers become unreachable, even when the query is for a site that's directly connected and reachable 💀
Strategies for implementing this architecture are discussed in appendix B 📖
To locate this "meta-domain" server inside a mid-level network, it's proposed that a nameserver entry for meta-dns exist within the mid-level network's domain 🔎
3.2 Public Domain Software 💿
File transfer traffic made up 23% of the NSFnet backbone traffic for May 1991. Public shareware is a super valuable resource on the Internet, and a lot of effort is going into building applications that track all the available resources in the public archives [SHHH90] 🗂️
It would be difficult, if not impossible, to create an up-to-date repository for every public domain package on the Internet. There's just too much software, and new stuff is being developed every single day 😵💫 Some hosts have become popular public archives (like uunet.uu.net, sumex-aim.stanford.edu, wuarchive.wustl.edu), and new developers tend to send their software to these sites as distribution points 📦 Maintaining centralized archives is also expensive, which is another reason people don't centralize. The UUnet archives at uunet.uu.net take up roughly 1GB of disk storage 💸
Recently though, a bunch of methods for resource discovery have been developed and are available on the Internet. Think the "ftp-list" file compiled by John Granose (odin@pilot.njin.net), Archie at archie.cs.mcgill.ca, and Prospero [NEU] 🧭
It's desirable for mid-level networks to provide up-to-date pointers to the distribution hosts for available public software archives. Coordinating the distribution of a static list is difficult (though not impossible), so using automated resource discovery mechanisms like Archie and Prospero is recommended 🤖 Under ideal conditions, software that is popular and significant (e.g., X11, TeX, RFC's) could be archived and distributed within the mid-level network. But measuring "popularity" and "significance" is debatable, so that's left for further evaluation 🤔 Also, a nameserver entry for host swdist within the domain can provide info on the different options for software distribution and discovery (static file location, pointers to Archie servers, etc.). This nameserver entry can be an alias for a CNAME or a TXT entry 🎁
3.3 Network Time ⏰
An important feature of any computer network that provides distributed services is being able to synchronize the local clocks on its various systems. Ideally, the clocks of all the reference sources would be synchronized to national standards by wire or radio ⚡ Network Time is super popular and important, which makes it a very useful potential service for a mid-level network to provide. No specific protocol for maintaining time is proposed here. Any available protocol that keeps time with reasonable accuracy could be used 🕰️
Network Time Protocol (NTP) traffic was 1% of the NSFnet traffic during May 1991. That might seem insignificant, but there have been times when a particular stratum-1 timeserver (for example, one of the stratum-1 servers at University of Delaware) got overloaded because too many different sites were trying to peer with it 🥵
It's proposed that at least one stratum-1 and two stratum-2 servers be located within a mid-level network. (Picking three servers is based on the NTP standards documentation [MIL89].) The servers can be at any of the directly connected sites in the network, as long as they're publicly accessible. All the sites connected to the mid-level network can then sync their system times with the servers inside the mid-level network itself 🫧 Besides making the timekeeping network more reliable, this setup also limits the load on each timeserver 🙌
For locating network time servers within a domain, nameserver entries for timekeeper-x (where x= 1,2,3..) can be made within the domain. The servers are numbered in order of preference and accuracy. So timekeeper-1.foo.net would be the primary timekeeper, and timekeeper-2.foo.net would be an additional (possibly secondary) timekeeper within the domain foo.net 🧚♀️ If there aren't any such hosts available within a domain, a TXT entry pointing to other recommended timeservers could be provided instead ⭐
Part 1 ends here, bestie. Part 2 is coming next with the rest of Section 3 and everything after it 🧚💖
3.4. Network News 📰
Okay so Network News (aka Usenet News) was 14% of the NSFnet traffic in May 1991 💀. Netnews is EXPENSIVE, bestie, in terms of disk and CPU power, plus all the network bandwidth it eats 😭.
The current setup is a bunch of hub sites spread across the Internet. End sites get their news feeds from other sites. To inject an article, you send it to your nearest "upstream" site, which forwards it to its connected news sites, and so on. It's giving distributed group chat 🫶.
There's no preset way to find a site willing to give you a feed, so it usually ends up being whichever site the admin happens to know 👀. That can totally backfire. Some sites can't get an economical feed from inside their own mid-level network and end up pulling it from a site on a different mid-level network. Not it 💅.
A mid-level network could fix this by providing a newsfeed to any or all of its directly connected end sites. It's still an expensive resource, but some of the costs can be moderated by acting as a transit news feeder, so the news doesn't have to hang around on disk for a long time 🧚. The software for the feed isn't specific and depends entirely on the newsfeed provider ✨.
3.5. Mailing Lists 💌
Internet mailing lists are another popular source of info, right alongside Network News 🌸. Like public software, though, there's no central repository of all the possible mailing lists on the Internet, and compiling one would take considerable effort. A fairly comprehensive list is available on the Internet and is mentioned in appendix A 📎.
There's no clear strategy yet for distributing or maintaining mailing lists. But sending mail to every individual end user directly can be very expensive for a site. If a clear strategy for maintaining a list of mailing lists can be devised, then mail exploders can be set up at the mid-level networks, and each one forwards the mail to exploders at the end sites. This reduces the load on the originating systems and gives a clean path for tracking down mailer problems 🦋.
Also, to stop bounced mail from propagating back to the originator of the message, the mailing lists should be set up so bounced mail goes to the "owner" of the list and not to the originator of the mail message 🙈.
A list of major mailing lists for the services discussed in this document is in appendix A 💜.
4. Experimental Testbeds 🧪
Okay, the mid-level networks are super good media for distributing new ideas and tech, because they have such good working relationships with their end sites and peer networks 🤝. Examples include the White Pages pilot project [RS90] established by NYSERnet, and the NSAP routing schema for OSI transitioning [CGC91], etc. 💅
The mid-level networks could set up cooperative experimental testbeds for testing and deploying new technologies like the ones above. Besides deployment and testing, this could also be a "help" service that gets end sites started with the new software 🎀.
The exact interaction between the mid-level networks here isn't very clear yet. Competition for members between them complicates it, so it needs more discussion 🫠.
5. Network Information Services 🗂️
There's a whole bunch of new and useful user services on the Internet that are hard to document, let alone list comprehensively 🤯. Some attempt has been made at documenting them [NNS], and a mid-level network can be the first point of contact for spreading that info on a wide basis 🌈. The info can be disseminated in a more controlled and complete way with this hierarchical approach, if each mid-level network keeps its directly connected sites up to date. NIC services also make the network easier and more attractive to end users 🧚. Examples are:
- provide information resources
- security advisory messages
- list of library catalogs [GL91]
- geographical information servers
- password generators
- resolve end user problems (user support)
These NIC services are discussed in detail elsewhere [SSM91] 📚. For accessibility info, an entry for "nic" could exist in the DNS for the domain (this could be a TXT entry listing email or phone number info for users or other NIC's) 📞.
6. Network Operations 🛠️
The Network Operation Centers (NOCs) at the mid-level networks need to team up to resolve network problems 🫶. If there's a problem between two mid-level networks, or an end site has trouble getting to any host, the mid-level network NOCs can be the initial point of contact 🚨. The procedures for interaction among NOCs and the formats for exchanging trouble-tickets are described elsewhere [JOH91, ML91].
It's important for cooperating NOCs to have contact info for their directly connected campus/organizational sites and for their peer mid-level networks 📇. A distributed mechanism for maintaining contact info could use a nameserver TXT entry for "noc", or maintain "finger" info for user "noc@domain" or "noc@noc.domain" 💖. A NOC "phonebook" listing contact info for the various NOCs works as a static, non-distributed mechanism. It's understood that the phonebook can contain outdated info, but the distributed mechanisms can provide correct and updated NOC info, provided the hosts are reachable at the desired time ⏰. If it's undesirable to publish the phone number or email address of the NOC for any reason, an entry saying "unpublished" (or words to that effect) could exist in the nameserver or "finger" entry instead 🙈.
7. References 📚
Here's the full reading list, bestie 🍓:
[BOG] Dunlap, K., and M. Karels, "Nameserver Operations Guide
for Bind Release 4.8", CSRG, Department of Electrical
Engineering and Computer Sciences, University of
California, Berkeley, California.
[CCI88] CCITT Blue Book, "X.500 Series Recommendations", ITU,
1989.
[CGC91] Collela, R., Gardner, E., and R. Callon, "Guidelines for
OSI NSAP Allocation in the Internet'', RFC 1237,
NIST, Mitre, DEC, July 1991.
[SSM91] Sitzler, D., Smith, P., and A. Marine, "Building a Network
Information Services Infrastructure", RFC in
preparation.
[GL91] George, A., and R. Larsen, "Internet Accessible Library
Catalogs & Databases", Aug 1991.
Available via anonymous FTP from ariel.unm.edu.
[JOH91] Johnson, D., "NOC TT Requirements", RFC in
preparation.
[MAN] Mandelbaum, R., and P. Mandelbaum, "The Strategic Future
of the Mid-Level Networks", University of Rochester,
NY, 1991.
[MOC87a] Mockapetris, P., "Domain Names - Implementation and
Specification", RFC 1035, USC Information Sciences
Institute, November 1987.
[MOC87b] Mockapetris, P., "Domain Names - Concepts and
Facilities", RFC 1034, USC Information Sciences
Institute, November 1987.
[MIL89] Mills, D., "Network Time Protocol", RFC 1129, UDel,
October 1989.
[ML91] Mathis, M., and D. Long, "User Connectivity Problems
Working Group", RFC in preparation.
[NEU] Neuman, B., "The Virtual System Model: A Scalable
Approach to Organizing Large Systems", Department of
Computer Science, University of Washington, FR-35,
Seattle, WA, May 1990.
[NNS] NSF Network Service Center, "Internet Resource Guide",
Cambridge, MA.
Available via anonymous FTP from nnsc.nsf.net.
[RS90] Rose, M., and M. Schoffstall, "The NYSERnet White Pages
Pilot Project", NYSERnet, Inc., Mar 1990.
[SHHH90] Schwartz, M., Hardy, D., Heinzman, W., and G.
Hirschowitz, "Supporting Resource Discovery Among
Public Internet Archives", Department of Computer
Science, University of Colorado, Boulder, CO.,
September 1990.8. Security Considerations 🔐
Plot twist: this memo just doesn't get into security at all 😭💀. Security issues are not discussed in this memo.
9. Author's Address 📮
Vikas Aggarwal
JvNCnet
6 von Neumann Hall
Princeton University
Princeton, NJ 08544
Phone: +1-609-258-2403
Email: vikas@jvnc.netAppendix A - Mailing Lists 💌
Here's a list of popular mailing lists for the services in this document 🎀. To subscribe to one, send a request to "mailing-list-request" (do not send a request to the entire mailing list, bestie 🙅♀️).
ietf@isi.edu: The general mailing list for the Internet Engineering Task Force. This group is concerned with the evolution and development of Internet-related protocols and standards. Old mail is archived at "venera.isi.edu" in directoryftp/irg/ietf🦋.ntp@trantor.umd.edu: For discussions on the Network Time Protocol (NTP) ⏰.namedroppers@nic.ddn.mil: Mailing list for discussions on DNS topics. Old mail is archived at "nic.ddn.mil" 🌸.
At the time of writing, a list of mailing lists on the Internet is available via anonymous FTP from host "ftp.nisc.sri.com" in the file "netinfo/interest-groups" 🧚.
Appendix B - DNS Architecture Strategy 🏗️
This section discusses practical strategies for implementing a nameserver architecture within a mid-level network, so that it can resolve nameserver queries for all domains directly attached to it 💅.
To resolve queries for all directly connected networks, a host that is authoritative for all directly attached domains will need to exist within the mid-level network. Nameservers at the end sites would then treat this "group-of-domains" nameserver as a forwarding server to resolve all non-local queries 🫶.
This can be done by adding a line to the named.boot file on the end-site nameservers, such as:
forwarders 128.121.50.7 128.32.0.4This method has the added advantage that the forwarding server builds up a very rich cache of data [BOG] and acts like a metacache that all hosts can benefit from ✨. Note that the forwarding server is queried only if the end-site server cannot service a query locally. Hence, the "meta-domain" server is not overloaded with queries for all nameserver lookups 👑.
📦 where this page comes from
- ietf: rfc1291 1991-12
- 💖 girlified by claude-haiku-5-5 on : a language model's rewrite of the original, under its terms