Internet naming models; Long message

Robert A. (Bob) Morris ram at CS.UMB.EDU
Fri Oct 12 15:35:15 CDT 2001


christian thompson writes:
 > Date:         Fri, 12 Oct 2001 08:48:18 -0400
 > From: christian thompson <cthompson at sel.barc.usda.gov>
 > To: TAXACOM at usobi.org
 > Subject:      Names, Numbers, "Agency buckaroos," etc.
 >
 > ...
 >
 > And for the proper interaction among systems, etc., there needs to be
 > central coordination. The Internet work only because once a small group of
 > US (="Agency buckaroos") computer people decided on how build such a system
 > to assign unique numeric addresses (which may also have ALPHA names too).
 > Yes, computer people surrendered some of their "freedom," but we all got a
 > very useful system.

Hannu Saarenmaa and I have been off and on working on a paper
discussing the utility for taxonomy of the models in both successful
and proposed internet naming schemes, of which there are several.

One of the interesting things about the Domain Name System (DNS)---the
alpha part of the naming system---is that delegation of naming
authority is hierarchical, not central. The top level domain names
(edu, org, com, gov, uk, ca, au, ...) are fixed by a central
authority. This authority maintains a central database of second level
domains (mit, usda, usobi, ibm, ...) but cedes to "domain registrars" the
authority to issue these names if they have not already been
registered. For each such issuance, the registering organization or
individual determines the entire hierarchy below it, including the
policies for authorizing further naming AND including delegation of
the entire authority for any subdomain. Thus, for example, the umb.edu
domain is administered by the University of Massachusetts at Boston
Computing Services, but they have delegated to the Computer Science
department the administration of cs.umb.edu. We have chosen not to
further subdivide, but we need no further authority or technical
infrastructure to do so. If we do subdivide further and follow the
technical rules of which you wrote, nothing whatsover breaks: the
subdivided names will be globally unique and all properly written
software will find the resource named. Similar principles are
in effect for the numeric part of the naming system.

There are several interesting issues with the DNS as a taxonomic
scheme and Hannu has someone thinking about some of them. To my mind,
the value of such an exercise is not to computerize taxonomy
successfully---laudable as that goal might be---but rather that
substantially more thought has gone into these issues in the internet
sphere than in the taxonomy sphere and there may be lessons to be
learned about naming schemes that have not occurred to taxonomists.

Curiously, some issues are not that different from what existing
taxonomic schemes suffer. They include: (a) A name can be
snytactically correct, reserved from further use, but not actually
refer to anything (b). The thing that the name refers to is itself a
reference to something else (an IP address), and both references may
be, and often are administratively changed. IP addresses are the
unique number sought by the number grail seekers, and their syntax and
assigment protocols are designed for software to easily locate the
computer or other object they name. They in turn point to something
that is approximately as unique as a U.S. Vehicle Identification
Number or Social Security Number and purports to never change. It is
physically associated with the network interface card of the computer
that is the thing named. Of course, that can be moved to a different
computer...).  Despite those changeable things, there is an effective,
distributed, software architecture (the Domain Name Server system and
the Internet routing system), which discovers the currently correct
target of those names and numbers. Think of it as a really good global
corps of taxonomic librarians with access to every document they need
to find what a name really refers to. (c). Duplication of names can
occur but are usually quickly discovered by the global system unless
the duplication is intententional and accompanied by technically
sophisticated schemes to hide the "spoof"(*) (d).Temporary inability to
find the resource that is named by the name can lead poor software
into believing that it doesn't exist, rather than that the software
hasn't done enough research. (e). Poorly written software can fail to
find the (usually emininently successful) fact of synonomy in the
taxonomic sense. That is, when names are changed, some systems may be
using the old names, usually leading to failure. There is a formal,
voluntary synonomy protocol (DNS "aliasing"), but nothing like a paper
trail of previous publications that allows tracking the history of
name changes, most of which are lost: the current name is highly
reliable, but previous names are virtually non-existent.  Most
interesting is that DNS names all have a time to live (TTL) in any
database holding them. Records are periodically and automatically
renewed---provided the database meets its responsibilities about the
TTL protocols.

For what it's worth, the Worldwide Web and its defunct predecessors
were the first internet facility to make people realize that on the
internet, unique identifiers were needed for reasons other than for
machines to reliably locate and communicate with each other. For the
most part, that realization has so far led mostly to standards about
syntax of names. The taxonomic issues remain a mine field in that
sphere. Interested people can do a google search on "URI" and choose
among the 982,000 hits---about twice as many as on "taxonomy".


Bob Morris

p.s.  All the biology agency buckaroos I've met since hanging with
this crowd have been dedicated, productive, and well informed
professionals, but I'm sure you know that the actual history of the
technical origin of the internet is that the buckaroos at the Defense
Advanced Research Projects Agency contracted its design and
construction to the cowboys and a (shamefully) few cowgirls at Bolt,
Baranek and Newman, Inc. and a handful of academic
organizations. However, some technical people did drift back and forth
between buckaroohood at DARPA and packet wrangling at BBN and the
universities.

(*)p.p.s.
Intentional and benign spoofing actually is in sudden widespread
consumer use in low-cost firewall gateways for cable modem and ISDN
users. It is a little interesting when thinking about
nomenclature vs identification. It is yet further afield, so I'll
shortly write down my thoughts on our web site and just announce it
here.




More information about the Taxacom mailing list