[14]            DNS: A Short History and a Short Future
                                    by Ted Byfield 
                                    tbyfield@panix.com

[Editor's Note: In the following article, Ted Byfield examines the
problem of domain naming in terms of the lessons from the experience
of the telephone.  His article presents the kind of broader
perspective that needs to be considered in trying to solve the
problems raised by the domain name system in the past few years.]

[Author's Note: This essay was first published on Rewired during the
week of 28 Sept 1998 under the title "A Higher Level of Abstraction";
I've slightly amended it for redistribution on nettime.  Thanks to
David Hudson for his excellent edit. TB]

     In the debates that have erupted over domain name system (DNS)
policy, two main proposals have come to the fore: a conservative
option to add a handful of new generic top-level domains (gTLDs:
".nom" for names, ".firm" for firms, etc.) administered by a minimal
number of registrars, and a more radical proposal to level the
hierarchical structure of domain names altogether by permitting openly
constructed names ("whatever.i.want") administered by an open number
of registrars.
     The supposed cause for these debates orbit around perceived
limitations on the system, monopolization of registration by NSI (in
the U.S., of course) and a scarcity of available names; as such, the
debates gravitate toward modernizing the system and preparing it for
the future.  What little attention has been paid to the past has
focused on the immediate past, namely, the institutional origins of
the present situation.
     Little or no attention has been paid to the prehistory of the
basic problem at hand: how we map the "humanized" names of DNS to
"machinic" numbers of the underlying IP address system.  In fact, this
isn't the first time that questions about how telecom infrastructures
should handle text-to-number mappings have arisen.  And it won't be
the last time, either; on the contrary, the current debates are just a
phase in a pas de deux between engineers and marketers that has
spanned most of this century.
     A bit of history: From the 1920s through the mid 1950s, the U.S.
telephone system relied on local exchange telephone numbers of between
two and five digits.  As these exchanges were interconnected locally,
they came to be differentiated by an "exchange name" based on their
location.  These names, two-letter location designations, made use of
the lettering on telephone keypads: thus an 86x-exchange, for
example, might be "TOwnsend," "UNion," "UNiversity," or "VOlunteer."
Phone numbers such as "UNion 567" were the norm; "86567" -- the same
thing -- would have been seemed confusing, in much the same way that
foreign dialing conventions can be.  There wasn't a precedent for a
purely numerical public addressing system, and, with perfectly good
name-and-number models like street addresses in use for centuries, no
one saw any reason to invent one.
     However, as exchanges became interconnected across the nation,
AT&T/Bell found a number of problems -- among them, that switchboard
operators sometimes had difficulty with accents and peculiar local
names.  As a result, the national carriers began to recommend
standardized exchange names, according to a curious combination of
specific and generic criteria: they chose words that resisted regional
inflection but were common enough to peg to "local" landmarks.  The
numbers 5, 7, and 9 were reserved because the keys have no vowels,
making it (so the theory goes) more difficult to form words from them;
hence artifacts like the fictional prefix 555, so common in old
movies, later became the national standard for prefix in fact, in the
form of directory assistance.
     By the late 1950s, when direct long-distance dialing became
possible, then popular, variable length phone numbers became a problem
for the national carriers, which demanded yet more standardization
seven-digit phone numbers in a "two-letter five-number" (2L5N) format.  
And while it wasn't an immediate problem, the prospect of international 
telephonic integration -- with countries that used different letter-to-
number schemes or even none at all -- drove yet another push for 
standardization, this time for an "all-number calling" (ANC) system.  
Amazingly, the transition to ANC in the U.S. took almost thirty years, 
up to around 1980 depending on the region.  (Just as certain telecom-
under-served areas are now installing pure digital infrastructures 
while heavily developed urban areas face complex digital-analog 
integration problems, phone-saturated urban areas such as New York 
were among the last to complete the conversion to ANC.)
     Direct long-distance dialing wasn't merely a way for friends and
family to keep in touch: it allowed businesses to deal in "real time"
with distant markets.  And the convention of spelling out numbers,
only partially suppressed, hence fresh in the minds of the many,
became an opportunity.  Businesses began to play with physical legacy
of lettered keypads and cultural habits by using number-to-letter
conversions as a marketing tool -- by advertising mnemonic phone
numbers such as "TOOLBOX." And as long-distance calls became a more
normal for people to communicate, tolls began to fall, in a vicious --
or virtuous, if you prefer -- circle, thereby lowering the cost of
transaction for businesses and spurring their interest in broader
markets.
     However, direct long-distance dialing presented a new problem,
namely the cost of long-distance calls, which became the next
marketing issue -- and toll-free direct long-distance dialing was
introduced.  The marketing game replayed itself, first for the
800-exchange (and again more recently for the 888-exchange).  As these
number spaces became saturated with mnemonic name-numbers, businesses
began to promote spelled-out phone numbers that were *longer* than the
functional seven digits (1-800-MATTRESS) -- because the excess digits
had no effect.  The game has played itself out in other ways and other
levels -- for example, when PBX system manufacturers adopted keypad
lettering as an interface for interactive directories which use the
first two or three "letters" of an employee's name.
     Obviously, this capsule history isn't in a literal allegory for
the way DNS has developed -- that's not the point at all.  There are
"parallels," if you like; questions of localized and systematic naming
conventions, of national/international integration, of arbitrarily
reserved "spaces," of integrating new telecom systems with installed
infrastructures, of technical standards coopted by marketing
techniques, and so on.  But implicit in the idea of a "parallel" is
the assumption that the periods in question are separate or distinct;
instead, one could -- and should, I think -- see them as *continuous*
or cumulative phases in an evolving effort to define viable standards
for the interfaces between "machinic" numerical addressing systems and
human linguistic systems.  Either way, though, DNS -- like the
previous efforts -- won't be the last, regardless of how it is or
isn't modified in the next few years.
     This isn't to dismiss the current DNS policy debates.  On the
contrary, they bear on very basic questions that should be addressed
*precisely because their implications aren't clear* -- questions about
national/international jurisdiction and cooperation, centralized and
distributed authorities, the (il)legitimacy of de facto monopolies,
and so on.
     Ultimately, though, these questions are endemic to distributed-
network communications and are *not* unique to DNS issues.  What *is* 
unique to DNS isn't any peculiar quality but, rather, its historical 
position as the first "universal" addressing system -- that is, a 
naming convention called upon (by conflicting interests) to integrate 
not just geographical references at every scale (from the nation to 
the apartment building) but also commercial language of every type 
(company names, trademarks, jingles, acronyms, services, commodities), 
proper names (groups, individuals), historical references (famous 
battles, movements, books, songs), hobbies and interests, categories 
and standards (concepts, specifications, proposals) ... the list goes 
on and on.
     The present DNS debates center mostly around the question of
whether and how DNS should be adapted to the ways we handle language
in these other spheres, in particular, "intellectual property." Given
the sorry state of that field -- which is dominated by massive
industrial pushes to extend proprietary claims indefinitely, to
criminalize infractions against those claims, and to weaken "consumer"
protections by transforming commodities purchases into revocable and
heavily qualified use-licenses -- it's fair to ask whether it's wise
to conform such an allegedly important system as DNS to that morass.
     What's remarkable is how quickly this has evolved, from a system
almost fanatically insistent on shared resources and collaborative
ethics to a speculative, exclusionary free-for-all.  A little more
history: With the erratic transformation of the "acceptable use
policies" (AUPs) of the various institutional and backbones supporters
of the Internet in the first half of this decade, commercial use of
the net expanded from a strictly limited regime (for example, NSFNET's
June 1992 "general principle" allows "research arms of for-profit
firms when engaged in open scholarly communication and research") to
an almost-anything-goes policy left to private Internet providers to
articulate and enforce (along with questions of spam, Usenet
forgeries, and so on and so forth).  The result was that any entity
that couldn't establish educational, governmental, or military
credentials was categorized as "commercial" by default.  The ".com"
gTLD quickly became the dumping ground for just about everything: not
just business names and acronyms, but product and service names
(tide.com, help.com), people's names (lindatripp.com), ideas and
categories (rationality.com, diarrhea.com), parodies and jokes
(whitehouse.com, tragic.com), and everything else (iloveyou.com,
godhatesfags.com).  (This essay omits discussion of the more nebulous
".net" and ".org" gTLDs -- which are vaguely defined and became
popular only after the domain-name debates -- as well as of state
[".ny"] and national [".uk",".jp"] gTLDs.) Thus, the
"commercialization" of the net took place on two levels: in the
legendary rush of business to exploit the net, obviously, but also in
the administrative bias against noninstitutional use of the net.
     There were practical reasons for that trend, to be sure:
individual or "retail" access was initiated by commercial Internet
providers, which doled out many more dial-up user accounts than
domains, as well as technical issues ranging from telecom pricing
schedules to software for consumer-level computers that discouraged
the casual use of domains.  But the trend also had an ideological
aspect: the entities that governed DNS preferred the status quo to
basic reforms -- and, in doing so, relegated the net's fast
diversification to a single gTLD that became less coherent even as it
became the predominant force.
     One can't fault the administrators for failing to foresee the
explosion of the net; and their responses are, if not justified, at
least understandable.  DNS was built around the structurally
conservative assumptions of a particular social stratum: government
agencies, the military, universities, and their hybrid organizations
-- in other words, hierarchical institutions subject to little or no
competition.  These assumptions were built into DNS in theory, and
they guide domain-name policy in practice to this day -- even though
the commercialization of the net has turned many if not most of these
assumptions upside down.  Not only are the newer "commercial" players
prolific by nature, but most of their basic assumptions and methods
are very much at odds with the idealized cooperative norms that
supposedly marked governmental and educational institutions: they come
and go like mayflies, they operate under the assumption that they'll
be besieged by competitors at any moment, they thrive on imitation,
and they succeed (or at least try) by abstracting everything and
laying exclusionary claim to every thing abstract -- procedures,
mechanisms, names, ideas, and so on.  The various systems and fields
we call "the market" worked this way before the net came along; small
wonder that they should work this way when presented with a "new
world."
     If no one anticipated the speed with which business would take to
this new medium, even less could anyone have predicted how it would
exploit and overturn the parsimonious principles that dominated the
net.  Newer domain users quickly broke with the convention of
subdividing a single domain into descriptively named sub- and
sub-sub-domains that mirrored their institution's structure (e.g.,
function.dept.school.edu).  Instead, commercial players started to
strip-mine name space with the same comical insistence that led them
to label every incremental change to a commodity "revolutionary." The
efficient logic of multiple users within one domain was replaced with
a speculative logic in which a few users became the masters of as many
domains as they could see spending the money to register.  In some
cases, these were companies trying to extort attention and money out
of "consumers" (business's preferred name for "person"); in other
cases, they were "domain-name prospectors" hoping to extort money out
of business; in many more cases, though, they were simply "early
adopters" experimenting with the fringes of a new field.  In effect,
the potentially complex topology of a multilevel name space was
reduced mostly through myopic greed and distorted rhetoric to a
flatland as superficial as the printed pages and TV screens through
which the business world surveys its prey.  The minds that
collectively composed "mindshare," it was assumed, couldn't possibly
grok something as complicated as a host name.
     So, for example, when Procter and Gamble decided to apply "brand
management" advertising theories to the net, it registered
diarrhea.com rather than simply incorporating diarrhea.pg.com into its
network addressing.  And so did the ubiquitous competition, including
the prospectors who set about registering every commercial domain they
could cook up.  The follies of this failed logic are everywhere
evident on the net: thousands of default "under-construction" pages
for domain names whose "owners"  renters hoping to become rentiers
wait in vain for someone to buy their swampland: graveyard.com,
casual.com, newsbrief.com, cathedral.com, lipgloss.com, and so on,
and so on.
     Under the circumstances -- that is, thousands of registered
domain names waiting to be bought out -- claims that existing gTLD
policies have resulted in a scarcity of domain names are doubtful.  
In fact, within the ".com" gTLD alone, the number of domain names
registered to date is a barely expressible fraction of possible domain
names, such as "6gj-ud8kl.com": ~2.99e+34 possible domain names
*within ".com" alone*, or ~4.99e24 domains for every person on the
planet; if these were used efficiently -- that is, elaborated with
subdomains and hostnames such as "6b3-udh.6gj-ud8kl.com" -- the number
becomes effectively infinite.
     Obviously, then, the "scarcity" of domain name is *not* a
function of domain name architecture *or* administration at all.  It
stems, rather, from the commercial desire to match domain names with
names used in everyday life -- in particular, names used for marketing
purposes.  To be sure, "6gj-ud8kl.com" isn't an especially convenient
domain name; but, then again, was "Union 567" or "+1-212-674-9850" a
convenient phone number, "187 Lafayette St #5B New York NY 10013" a
convenient address, or "280-74-513x" a convenient Social Security
number?
     But if DNS is in fact such an important issue, does it really
make sense to articulate its logic according to the "needs" of
marketers? After all, business has managed to survive the tragic
hardship of arbitrary telephone numbers for decades and arbitrary
street addresses for centuries.  Surely, if the net really will
revolutionize commerce, to the point of "threatening the nation-state"
as some like to claim, the inconvenience of arbitrary domain names
will hardly stop the revolution.
     *Of course* there are territorial squabbles over claims to names
and phrases.  And *of course* some people and organizations profit
from the situation.  But we don't generally erect a stadium in areas
where gang fights break out; so one really has to ask whether it's a
good idea to restructure gTLD architecture -- supposedly the system
that will determine the future of the net, hence a great deal of human
communication -- to cater to a kind of business dispute that's in no
way limited to DNS.
     Ultimately, it doesn't really matter which proposed gTLD policy
reform prevails, because the gains will be mostly symbolic, not
practical -- except, of course, for the would-be registrars, for whom
these new territories could be quite profitable.  At minimum, adding
new gTLDs such as ".firm", ".nom", and ".stor" will bring about a few
openings -- and, more to the point, a new round of territorial
expansions, complete with redundant registrations, intellectual-
property lawsuits, etc.  At maximum, an open domain-name space that 
allows domains such as "what-ever.i.want" will precipitate a domain-
grabbing free-for-all that will make navigating domains as
unpredictable as navigating file structures.
     Moreover -- and *much* worse -- where commercial litigation is
now limited to registered domain names, an open namespace would invite
attacks on the use of terms *anywhere* in an address.  Put simply:
where apple.material.net and sun.material.net are now invulnerable to
litigation, in an open namespace Apple Computers and Sun Microsystems
could easily challenge "you.are.the.apple.of.my.eye" and 
"who.loves.the.sun".
     Neither proposed reform *necessarily* serves anything resembling
a common good.  But both proposed reforms will provide businesses with
more grist for their intellectual property mills and provide users
with the benefits of, basically, vanity license plates.  The net
result will be one more step in the gradual conversion of language --
a common resource by definition -- into a condominium colonized by
businesses driven by dreams of renting, leasing, and licensing it to
"users."
     It doesn't, however, follow that the status quo makes sense -- it
doesn't.  It's rife with conceptual flaws and plagued by practical
issues affecting almost every aspect of DNS governance -- in
particular, who is qualified to do it, how their operations can be
distributed, and how democratized jurisdictions can be integrated
without drifting being absorbed by the swelling ranks of global
bureaucracies.  The present administration's caution in approaching
gTLD policy is an instinctive argument made by people happy to
exploit, however informally, the *superabundance* of domain-name
registrations.
     Without doubt, the main instabilities any moderate gTLD policy
reform introduced would be felt in the administrative institutions'
funding patterns and revenues.  More radical reforms involving more
registrars would presumably have more radical consequences, among them,
a need to certify registrars and DNS records, from which organizations
with strong links to security and intelligence agencies (Network
Associates, VeriSign, and SAIC) will surely benefit.  The current
administration insists that an open name space would introduce
dangerous instabilities into the operations of the net.  But whether
those effects would be more extreme than the cumulative impact of
everyday problems -- wayward backhoes, network instabilities, lazy
"Netiquette" enforcement, and human error -- is doubtful.
     There is one point on which the status quo *and* its critics
agree: the assumption that DNS will remain a fundamental navigational
interface of the net.  But it need not and will not: already, with
organizations (ml.org, pobox.com), proprietary protocols (Hotline),
client and proxy-server networks (distributed.net), and search-engine
portal advances (RealNames, bounce.to), we're beginning to see the
first signs of name-based navigational systems that complement or
circumvent domain names.
     And they're doing it in ways that address not the bogeys that
appear in the nightmares of rapacious businessmen but the real
problems and possibilities that many, many more users are beginning to
face: maintaining stable e-mail addresses in unstable access markets,
maintaining recognizable Zine-like servers in the changing conditions
of dynamic IP subnets, cooperating under unpredictable load
conditions, and, of course, *finding* relevant info -- not *offering*
it, from a business perspective, but *finding* it from a user's
perspective.
     DNS, as noted, was built around the assumptions of a specific
social stratum.  Prior to the commercialization of the net, most users
were if not computer professionals then at least technically
proficient; and the materials they produced were by and large stored
in logical places which were systematically organized and maintained.  
In short, the net was a small and elite town, of sorts, whose denizens
-- "Netizens" -- were at least passingly familiar with the principles
and practices of functional design.  In that context, just as multiple
users on a single host was a sensible norm, so were notions of
standardized file structures, naming conventions, procedures and
formats, and so on.  But just as the model of multiple users on a
single host has become less certain, so has the rest.
     The net has become a non-systematic distributed repository used
by more and more technically incompetent users for whom wider
bandwidth is the solution to dysfunctional design and proliferating
competitive formats and standards.  Finding salient "information" (the
very idea of which has changed as dramatically as anything else) has
become a completely different process than it once was.
     This turn of events should come as no surprise.  As commercial
domains multiplied, and as users multiplied on these domains, the
quantities of material their efforts and interactions produced grew
ferociously -- but with none of the clarity typical the "old"
institutional net.  In the past, the information generated around or
available through a domain (or to the subdomains and hostnames
assigned to a department in a university or military contractor) was
often "coherent" or interrelated.  But that can't be said of the
material proliferating in the net's fastest-growing segments:
commercial Internet access providers, institutions that automatically
assign Internet access to everyone, diversified companies, and any
other domain-holding entities that permit discretionary traffic.
     Instead, what one finds within these domains is mostly random
both in orientation and in scale: family snapshots side by side with
meticulously maintained databases, amateur erotic writings next to
source-code repositories, hypertext archives from chatty mailing lists
beside methodical treatises, and so on.  In such an environment, a
domain name functions more and more as an arbitrary marker, less and
less as a meaningful or descriptive rubric.
     This isn't to say that domain names will somehow "go away"; on
the contrary, it's hard to imagine how the net could continue to
function without this essential service.  But the fact that it will
persist doesn't mean that it will serve as a primary interface for
navigating networked resources; after all, other aspects of network
addressing have become all but invisible to most users (IP addresses
and port numbers to name the most obvious).
     The benefit that DNS offers is its "higher level of abstraction"  
a stable addressing layer that permits more reliable communications
across networks where changing IP numbers change and heterogeneous
hardware/software configurations are the norm.  But "higher" is a
relative term: as the substance of the net changes as what's
communicated is transformed both in kind and in degree, and as the
technical proficiency of its users drops while their number explodes
DNS's level of abstraction is sinking relative to its surroundings.

---------------------------------------------------------------------
Reprinted from the Amateur Computerist Vol 9 No 1 Winter 1998-1999. 
The whole issue or a subscription are available for free via email. 
Send a request to jrh@ais.org  or see  http://www.ais.org/~jrh/acn/
---------------------------------------------------------------------