Truly common trees have one single root, but how many roots the forest? (BTW some trees have more than a single root. Some even have no root). The forest is the international data network name space, de facto created in 1977 at FCC's initiative, and then under FCC and States controlled or direct monopoly interconnect agreements, within the frame of ITU (CCITT at the time), ie the "Intlnet" as sometimes nicknamed. That "forest" delegated the ARPANET sub-namespace ".arpa" in 1983/84 with provisions for the ccTLD support, possibility to extend to new sub-namespaces (multiorganization TLDs), then current root "bouquet" interconnect (com, net..) plus its own ones (edu, gov, mil..), resulting into the difference between internal sub-namespaces administration (legacy) and related sub-namespace registration (ccTLDs and foreign nets) to avoid conflicts. This led to the reversed order, right to left on dot separated labels, and to the need of the final dot, as a reversal indicator (also understood as the global root indicator), to avoid confusion with the ITU X121 and E.164 namespaces. The resulting constraints and possibilities for the Internet have been fully documented by Jon Postel in the RFC 920 (Oct 1984). This insured a clean and secure management of the ".arpa" sub-namespace with no conflicts with its neighbors during 18 years - until ".biz" and the ENUM eventual support (it was an Internet non directly supported part of the 1984 deal). An exegesis of RFC 920 is a must for a namespace understanding. It is true that those who were no part to that "delegation" process (then perceived as a standard private network gateway tuning) only see the result. All the more than the Intlnet system went X121 permitting less developed technologies to digitally support numeric names (X.25/75/E.164). A confusion is also the wrong use of the "alternate roots" wording for extended copies of the ".arpa" sub-namespace root which focuses the debate on side cross-root possible conflicts instead of the considering the whole namespace issue. A root name - as coined at the CCC (common carrier club) in 1977 - is the mutually registered name of a sub-namespace into the global namespace (creation) and the other sub-namespaces (interconnect): ex. ".arpa", E.164, ".fra", ".com", ".net", etc.. The confusion we often meet is that the ".arpa" tree hides us the forest. And we confuse the interconnect roots with the main root. Let think to water trees species with multiple roots. The problem we face today is that many people, organizations, states have legitimate rights and,more importantly, have needs for their own sub-namespace with the importance the technology development permitting to evade from digital oriented addressing restrictions, and the importance taken by the brainware (ways the users use a system), the enlarged common use of the IP networks all over the planet and the availability of a common shared IP network infrastructure. The DNS is a good candidate to assume a global namespace unified "supervision". But the DNS has important architectural/operational lacks (management, backup, security, stability, inter network technologies support, capacity, lack of extended services hooks, etc.). IMHO the question are: 1. is the DNS to be the universal global root (forest) URI resolver? Or a new "real names" by MS? 2. if yes what are the DNS+ and DNS.2 specifications to be? 3. iteratively what are to be the utilization architecture specifications to fully take advantage from them? 4. what is the impact on the URI semantic? 5. as the URI is the network (and our entire world) command line, what are iteratively the impact on software/brainware, technical opportunities, specifications for open network wide distributed processing? IMHO these are questions worth to investigate and to work on? Sorry to have been long on this. jfc At 08:50 29/07/02, Brian E Carpenter wrote: >Keith Moore wrote: > > > > > Keith Moore <moore@cs.utk.edu>: > > > >"alternate DNS roots" aren't part of DNS. if someone wants to propose > > > >a URN based on a DNS-like system with its own root zone, they're free > > > >to do so and see if they can get support for it. For that matter if > > > >someone wants to propose a URN based on some other naming system that > > > >doesn't look like DNS they're free to do that also. > > > > > > > >But trying to make "alternate DNS roots" fit into a DNS URI scheme is > > > >like trying to make OIDs or some other naming scheme fit into a DNS > > > >URI scheme. We don't need to do that - there's a separate scheme for > > > >OIDs. And trying to do so would make DNS URIs far more complex than > > > >they need to be - for no real benefit. For instance, how do you > > > >assign names to the alternate roots? > > > > > > By specifying the root name as a prefix? > > > > great. then people can start arguing about who gets to maintain the > > set of names for ... er... what were formally known as roots. > > most of us have better things to do with our time. > > > > > I agree that alternate roots are not part of DNS as long as you > > > contrain your universe to be the ICANN/USG published set of DNS > > > names, but there are other things floating around the net that > > > do use the DNS protocols and do resolve names for people who > > > choose to use them. > > > > there are other protocols on the net than those defined by IETF > > standards, too. the fact that they exist does not compel IETF to > > endorse them. > >Amusingly, Richard has, by suggesting that we should name the >"alternate roots", just discovered why the whole "alternate root" >story is nonsense. > >Since the DNS is a hierarchical namespace, and since trees have one >root, if you add "alternate roots", you then discover that you have >to uniquely name them, i.e. insert a new unique root "above" the various >"alternate roots". > >Or to put it another way, if we need several naming authorities, one for >each "alternate root", we're going to need a naming authority to uniquely >name those naming authorities. > >Funny how you can't change mathematical facts, isn't it? > >However, back in the real world, the existing unique root works just fine. > >Brian