Re: [wg-c] Re: nine principles for domain names

[Dave Crocker <dcrocker@brandenburg.com>]

At 04:35 PM 2/21/2000 -0500, Milton Mueller wrote:
> > "We believe that the requirement that all gTLDs have associated
> > descriptions of their intended meaning."
>Bret Fausett has just decisively answered this question. It is not only 
>possible,
>but highly likely, that the most popular new TLDs will be ones that are 
>notable
>only for their memorability, not for their "meaning." E.g., what is the 
>meaning of

It is always fascinating to see definitive statements about the future, 
particularly when there is no objective basis for treating the statements 
as anything other than pure conjecture.

As to "memorability" vs. "meaning"...  This attempts to draw a stronger 
distinction than the context allows, especially since this context is not 
one limited to psychologists and/or linguists.  The distinction also might 
be wrong.

That is, this discussion group is not one for highly precise linguistic 
distinctions, absent careful definition and group agreement.  One of the 
sources of conflict and confusion in these discussions has been from one 
person, or another, trying to assert a particular, precise definition 
(legal, technical. or the like) when the group, overall, is using the term 
more loosely.

Then we get to the question of "memorability".  Many things aid the ability 
to remember.  Semantic association is one of them, and a particularly 
strong one for "strings".  I suspect that the problem, here, is an attempt 
to treat strings that do not appear in a dictionary as having no "meaning" 
but possibly having some memorability.  That is a far too precise and 
constraining a distinction for this discussion (or just about any other, in 
my opinion.)  "XXX" is not in a dictonary but has nearly globally-clear 
meaning.

The modified text I suggested calls for a gTLD to have some associated text 
which describes its intended use.  The association between that text and 
the gTLD string is the intended "meaning".  The presence of that meaning is 
intended to underscore the mnemonic potential for the gTLD.

I suspect what Milton is forgetting is that all strings with a "meaning" do 
not have it inherently.  It is through association with other things in the 
world that we impart that meaning.  Hence, there is nothing in the least 
wrong, or even unusual, to refer to the string "zzz" as having meaning, 
given a legitimate associative context.


>.zzz? Who would want to register there? No one knows for sure, but I feel very
>confident that such a TLD would end up with a lot more registrations than the
>plonky .store proposed by IHAC. Same goes for a string of Chinese 
>characters, which

Milton's confidence is nice, but unfounded.  That does not mean he is wrong 
-- though *I* am confident he IS...

More importantly is that the interesting question is not whether it would 
be more popular, but why it would be. I am confident that a serious effort 
to learn the answer will disclose it to be that the string has a particular 
MEANING for its target audience.


>may evoke several meanings.

More than one meaning is just fine.  There is nothing wrong with a gTLD 
having the encoding efficiency of poetry.


> > What makes the small group of registry operators better than other bodies
> > for doing the selecting?
>
>The simple fact that it is their investment at risk. They are bidding 
>their labor
>and their money that people will want to register under a specific name. 
>Registry

Right.  Like NSI, using a granted monopoly.

Sorry, but they are plugging into an inherently restricted environment and 
an environment which must display broad-based stability. That is, stability 
is not just a question of whether servers respond, but whether the entire 
DNS operations and administration cycle is stable.


>operators are directly accountable for what they offer in a way that ICANN and
>certainly this little working group can never be. If we make stupid or unduly

NSI is "accountable" to its shareholders, not its users.  It pretty much 
has ignored is customers, with improvements coming only as its monopoly 
control over them has been threatened.  That threat came from controlling 
authorities, not "the market place."


>restrictive decisions, what are the consequences to the members of this 
>committee?
>There are none. It doesn't affect our paycheck or our workload in any 
>significant
>way. If on the other hand a registry offers services that no one wants, 
>there is a
>substantial loss. This is true regardless of whether the registry is for 
>profit or
>not.

That's fine, if you only look at things from the perspective of the registry.

It's not so fine if you look at it from the perspective of the customer.


> > Anticipating a "let the market decide" line of thinking will get us to
> > companies' experimenting with the marketability of a string and then going
> > out of business when it is not profitable.  At that point, the unfortunate
> > organizations that chose to register under the TLD will be left without
> > service.
>
>Several fallacies here. First, zone files can easily be escrowed and 
>transferred to
>another operator. Any contract between ICANN and TLD registries can be 
>structured

That's like the cliche that fills the upper-left and lower-right of a board 
with equations and ties them together with the phrase "let a miracle 
happen."  The suggestion sounds easy, because talking about it IS easy, but 
the details to make it happen are rather more difficult.

World peace is easy.  It only requires a simple agreement and a simple 
enforcement mechanism.

In addition, the hand-wave includes a technical error:  zone files do not 
contain customer record information.  If this registry failure-handling 
mechanism is to work, it must transfer all relevant customer data, not just 
what is in the zone file.

So the word "easily" is not quite so easy to apply, here.


>to minimize this risk and to compensate for it when it happens.
>
>Incidentally, similar risks occur if your ISP goes out of business, but I 
>don't see

ISPs rarely have any sort of monopoly control over end-users, since the 
cost of changing ISP is typically small.  For that matter, one reason to 
have a domain name is to reduce the ISP's control, by dis-associating the 
customer from the ISP's own domain name.

I've changed ISPs a number of times, but it has been transparent to those 
dealing with me.  That would not have been true if I were forced to change 
my domain name.


> > This gives a good indication of the reason that DNS service needs to be
> > counted as an infrastructure and the sole-source technical constraint --
> > only one organization can be registrar for a given TLD -- mandates that
> > operational stability included assurance that TLDs persist.  Going out of
> > business is an extreme form of instability.  We do not tolerate it for
> > telephone numbers or street addresses; why tolerate it for domain names?
>
>This is a regulatory issue that can be and is handled at the national 
>level. At the

Huh?  gTLDs are not "national" entities.  They are global.


>national level, we have professional regulators functioning according to
>established laws with significantly more resources than DNSO. Telephone 
>numbers are
>unique globally, as are domain names, but the ITU committee that 
>administers the
>numbering plan doesn't license all world telephone companies, nor attempt 
>to tell

In the case of telephone numbers, the ITU delegates to national 
agencies.  They, in turn, delegate number space below them.  So the ITU 
very much DOES control allocation of the first-level portion of the number 
plan.

In the case of gTLDs, ICANN does NOT delegate to national agencies.  It 
does need to delegate to some set of agencies.  And it needs to entail 
control on a par with what the ITU does for telephone country codes.


>them how to run their business. It is simply a coordinating agency. That 
>is ICANN's
>only legitimate role.

If you think the ITU is only a "coordinating" agency for its assigned areas 
of responsibility, I suggest you would greatly benefit from spending time 
in some ITU committees...


> > [Re: numbers of new TLDs] High estimates are VERY high.  Low estimates 
> are in the
> > thousands.  If we believe the low estimates and they are wrong, we can 
> add more
> > names
> > safely.
>
>OK. Let's start with thousands, then instead of millions. See, Dave, I'm 
>always
>willing to compromise. And since the largest list of proposed TLDs I've 
>seen does

Milton's declared flexibility notwithstanding, an upper-limit of 
"thousands" has been the accepted (lower) reference for several years.


>not exceed 300, we can conclude that there is no need to make restrictive 
>choices

Interesting number, 300.  That was the preliminary limit in the Postel 
proposal.  So the current list would consume ALL of the space originally 
proposed.

In any event, taking a current list of 300 and somehow extrapolating low 
demand for names from it suggests a very different approach to system 
management (and market projection) than any responsible operator would take.

A responsible operator must anticipate greater growth than is "obvious" 
since they cannot afford to have too little capacity.  They must make 
estimates based on high-end possibilities, asking "what if" things grow 
faster than they have planned.


>as to which ones are allowed and which are not. Scarcity in the economic sense
>exists when demand exceeds supply. Your lowest estimate vastly exceeds known
>demand.

Let's see.  This "known demand" is based on an operation that had no growth 
(no new gTLDs) since the net became a consumer market.

So there is no objective basis for guessing what will happen if the 
floodgates are opened without any controls on them.

That's a very interesting way to manage a critical piece of 
infrastructure.  Just a tad more fragile than a responsible operator would 
indulge in.


> > If we believe the high estimates and they are wrong, we break the
> > DNS (and, therefore, make the Internet useless for almost all users.  Hence
> > there is a requirement to approach the question of name space size with
> > significant caution.  It's fine to call that constraint "artificial" but
> > the label does not make the constraint any the less advisable.
>
>According to your comments above, you consider a few thousand new TLDs as 
>being on
>the extreme low end of caution. Since this WG is proposing to start with 
>6-10 and
>add on more at the rate of 50-100 a year, this is really a non-issue. We 
>are being
>cautious to a fault.

Only if one constrains their concerns to the technical, name-space capacity 
of the root.

Remember that "stability" entails a broader set of issues, including DNS 
administration.  For THAT, alas, we have a very poor expansion history -- 
as in none -- for gTLDs.



At 03:36 PM 2/21/2000 -0800, Karl Auerbach wrote:

>This layer of indirection is the main virtue of DNS.  The fact that humans
>have (temporarily) been involved in typing in DNS names, which has caused

Slipping in the parenthetical (temporarily) presumes a change in computer 
and network use that has been projected for nearly 20 years.  It has yet to 
occur and there are good reasons to doubt that it ever will.  Business 
cards are one example.  You need to be able to put useful strings onto 
business cards.  The holder of the card must be able to use the string.

Perhaps everything will be entirely automated in the future, but that is 
not going to happen for the global Internet anytime soon.

Changes to infrastructure mechanism happen slowly.  Changes to a global 
infrastructure happen VERY slowly.


>all of this debate, is merely an ancillary characteristic of DNS.  And the

Ancillary, in Karl's opinion.  Others view the mnemonic feature of domain 
names to be an extremely important benefit.

Remember that hostnames were introduced at a time when the user population 
was a small set of geeks who were comfortable debugging programs in 
octal.  That community thought dealing with numbers was just fine, so there 
must have been some reason the layer of abstraction was chosen to be 
human-oriented "semantic" strings.  That reason was not minor then and is 
not minor now.


>end of that human relationship with DNS is something that will slowly and
>substantially diminish as true directory services become more of a true

Again, please note that that claim has forcefully been put forward since 
the beginning of the X.500 directory service standards effort, roughly 20 
years ago, and it is not yet close to fruition.


>And when one says "stand for something" one is saying something about the
>linguistic context of a human user.  And given the billions of people on
>this planet who speak non-European languages, any choice of linguistic
>context is necessarily going to be the wrong choice for a significant part
>of the world's population.

1.  And it will be RIGHT for a significant part.

         And there is nothing horrible about having some disparity there.

2.  People are pretty good about acquiring and using "foreign" terms.

         The Malaysian term for table is "meja"; that's Portuguese.  An 
American term for "a little bit" is skosh; that's Japanese.  And then there 
is the global term, Coca-Cola...

         The fact that a DNS string might not be immediately intuitive to a 
user segment is not an automatic impediment.  There is a yin and yang 
balancing act when choosing names...

         Although there is merit and convenience in Winer's observation 
about the global prevalence of English, I do not believe that point is 
essential to this issue.

d/

=-=-=-=-=
Dave Crocker  <dcrocker@brandenburg.com>
Brandenburg Consulting  <www.brandenburg.com>
Tel: +1.408.246.8253,  Fax: +1.408.273.6464
675 Spruce Drive,  Sunnyvale, CA 94086 USA