brock@pi.cs.unc.edu (J. Dean Brock) (07/05/90)
philipp@GIPSI.GIPSI.FR (Philippe Prindeville) writes: > >... Sort of like the UC mess, >Berkeley.Edu instead of Berkeley.UC.Edu, etc. You're getting close to a touchy point with many American state universities. For example, unc.edu is actually the domain of University of North Carolina at Chapel Hill. The EDU domain already has four other registered institutions (unca.edu, uncc.edu, uncg.edu, uncwil.edu) whose official names are of the form "University of North Carolina at *". (Plus there's uncecs -- University of North Carolina Educational Computer Service.) Why are they not all of the form *.unc.edu? (Or the California ones of the form *.UC.edu?) Well, the universities themselves are really quite independent. There is an administrative body called the "University of North Carolina" but the universities don't consider themselves branches of that administrative body. Why does one university call itself unc.edu, rather than uncch.edu? Most state university system started with one campus founded many, many years before the others. The Chapel Hill campus was founded in 1789 (making it one of three state universities claiming to be the oldest in the United States). After a hundred plus years of being the only University of North Carolina, people start to assume the one in Chapel Hill is the default. Also, it's alumni are quick to point out that UNC-CH is THE UNC. (And even quicker to argue with South Carolinians about what university is THE Carolina. Oh, the other one once had the nerve to use the "carolina" alias in the CSNET.) And, finally, UNC-Chapel Hill was the first one of the UNC-*s to register and that's what really counts -- right? However, it is sometimes difficult to map the common name of the university to the domain name, uiuc.edu was one with which I had troble. Is domain registration always first-come-first-served? Are there any rules, either written or informally enforced?
loverso@Xylogics.COM (John Robert LoVerso) (07/06/90)
> Is domain registration always first-come-first-served?
Yes. Years ago I registered the domain name for SUNY at Buffalo,
and asked for Buffalo.EDU. However, someone at the time told me
to ask for "NewYork.EDU", simply because "it's not taken yet". It
would be utterly confusing had I done something [so stupid].
Before going the "Buffalo.EDU" route, I had tried to organize a
"SUNY.EDU" domain, but got little response from the other universities
and colleges in SUNY. The end result is today the various schools
are listed as "SUNYSB.EDU" or "Albany.EDU".
John
--
John Robert LoVerso Xylogics, Inc. 617/272-8140 x284
loverso@Xylogics.COM Annex Terminal Server Development Grouppjg@urth.acsu.buffalo.edu (Paul Graham) (07/07/90)
zben@UMD5.UMD.EDU (Ben Cranston) writes: |We won't even talk about eng.umd.edu (our "Engineering" subdomain) or our |cs.umd.edu (Computer Science Department) subdomain. My personal hope is |that once the patina fades and the magnitude of the continuing effort to |MAINTAIN these things becomes institutionally evident that domains will |re-devolve back to organizations geared to do first class management. bzzzzt. the dns is distributed. this means that if your hostmaster is lax you punish yourself. if the parent domain hostmaster is a lout you get to take care of yourself. or even less maliciously if the parent hostmaster is simply overworked you have some control over your namespace. it's been my experience that it's most often those at the top that feel the need to exert control (so often in the name of "first class management").
sob@TMC.EDU (Stan Barber) (07/08/90)
While the DNS is distributed, there remains a big problem with the in-addr.arpa domain. While eng.umd.edu could be handed over to someone to operate, it does not necessarily follow that 8.128.in-addr.arpa could be handed over. If the enterprise network (or facility network if you like) is subnetted and the entire subnet can be operated by one administrative group, it could work. Otherwise, it won't. I would guess that there are two major problems at many sites: 1. There are very few people who understand how to run the DNS right at any one site. Those sites with computer science departments and the like may have a larger number than some, but generally speaking the number is still low. 2. Unless growth of departments have been well thought out or enough money exists to fully subnet as needed, there will sometimes be more than one administrative group on the same wire and that wire may be a single legitimate subnet of the enterprise network. If the DNS is distributed within the main organization, who runs the in-addr.arpa for the subnet? [I am aware of the hacks in many routers that allow more than one subnet on a wire. What if you can afford to have fancy routers? Say, you can only afford the PC Routers or using a Sun as a router. What do you do?] There is also the question of ultimate responsibility. Who is ultimately responsible for the DNS run by a particular organization? What if that person is not empowered to correct problems with the DNS in the sub-domains of that organization? To me, the way to correct this problem is to create appropriate procedures to allow subnetwork managers to register hosts with the central authority for the site. That central authority would then be accountable for insuring the accuracy of the information and timely updates to the central databases. If the central authority is overworked, then funds need to be allocated by the organization to improve the service provided by the central authority. It is not a question of power. It is a question of good service, accuracy of data, and cooperation.
roode@ORC.OLIVETTI.COM (David Roode) (07/08/90)
There is a dichotomy between DNS management and address assignment. The in-addr.arpa domain is tied to the scheme in use on a particular level 2 network for address assignment. As you point out, through use of IP subnets the host address assignment can be distributed in a fashion similar to the DNS. IP subnets also distribute network management duties. There is no fixed mapping between the respective decompositions of IP nets and domains. You can have 4 domains on 1 IP subnets, 6 IP subnets in 1 domain, etc. You can also have instances of 4 domains on each of 10 subnets. This is a powerful model and I think it is not good to argue to institute a fixed mapping scheme just for the sake of in-addr.arpa. Whatever scheme is used in each IP network for address assignment and in each domain for subdomain creation has to take into account the realities of that site. One answer is not right for every site. The designs are explicitly specified to permit this flexibility.
sob@TMC.EDU (Stan Barber) (07/08/90)
While flexibility is generally a good thing, some combinations don't work well in the real world. DNS management and address management are interwoven to some extent. If the two are not coordinated, things don't work. Certain people have already complained that the lack of a specific requirement by the NIC on the assignment of in-addr.arpa (verus the rather explicit requirements for the assignement of a domain) already cause problems with address mappings to hostname (e.g. gethostbyaddr()). Imagine if the in-addr.arpa and the DNS for some group was controlled by two different organizations. Is that not problematical? I submit that it is. There are some organizations where this can be demostrated. STAN