[wg-c] lock-in

[Kent Crispin <kent@songbird.com>]

                  URL Lock-in and Competition

In a recent message I remarked that the requirement for large TM
owners to register their names in many TLDs is a form of "lock-in", a
form that operates in addition to the more widely discussed lock-in
caused by links in web pages.  Given the questions that have come up
in recent days, it seems useful to review that issue.  It is hard to
know how much background people have, so I am repeating stuff that
will be familiar to many old-timers, but may not have been apparent
to more recent arrivals. 

The term "lock-in" is used generically to describe a situation where
the sale of a particular commodity or service requires the customer
to return to the same vendor for futher service, because not to do so
would cause large additional costs.  IBM's use of proprietary
interfaces and protocols, many eons ago, is an example of such
mechanism: they implemented specialized interfaces for peripherals,
such that if you bought a computer from them, you were also obliged
to buy all your peripherals from them.  If you further invested in
development of software, or had large volumes of data stored in
EBCDIC, you faced additional costs if you changed vendors. 

Lock-in is good for the vendor, bad for the customer, and bad for
society as a whole.

                        Competition

It is worth noting that the success of the Internet stems in part
from the fact that proprietary protocols have been shunned -- the
non-proprietary internet protocols serve as structure where extremely
robust competition can thrive, and that is true in part because the 
non-proprietary protocols prevent lock-in. 

That is, central community control over standards resulted in far 
more competition, and ultimately more social good, than having 
individual companies develop competing protocols.

This is a particular case of a general point about competition -- 
competition requires structure, a set of common rules -- before it 
can be effective.  However emotionally satisfying, assassinating 
your competitors is discouraged...

This raises a point of critical importance in the debate in wg-c over
business models.  It is not a debate about the desirability of
competition -- everybody agrees that competition is useful.  It is
instead a debate about the rules that should govern that competition.
More on that in a latter message.


                          URL Lock-in

"Lock-in" in the context of web links (URLs) stems, fundamentally, 
from the one-directional character of web links -- the referenced 
web object has no information as to the identity of the referring 
object.  

Links are contained in bookmarks, individual web pages, and in search
engines.  Alta Vista is reported to index about 16% of all web pages;
a query of the form: "link:amazon.com" returns "About 601,068 pages
found." A ball-park extrapolation, then, is that there are more than
3 million pages with links to amazon.com.  

Those pages are not under the control of Amazon, and more
importantly, there is no way for Amazon to contact the maintainers of
those web pages.  In all probability, the majority of those pages are
not even actively maintained, and will never be changed.

For the vast majority of the entities behind those links, the link is
the only information that the entity has concerning the location of
Amazon.  They don't have a phone number or a street address, and even
if they did, they would have very little incentive to use them.  Note
further that those links were developed primarily through the medium
of the web -- Amazon does other forms of advertising, of course, but
it built its business through online communication.

Every one of those 3 million links, of course, contains the domain
name "amazon.com".  The domain name, for practical purposes, is the
companies business location in cyberspace.  If the domain stops
working that business location simply winks out of existence, and
Amazon's primary means of communicating with customers, its web site,
vanishes at the same time. 

So, if the domain "amazon.com" stops working, Amazon's business
stops, totally, and it can't tell its customers why, or what they
could do to fix it (such as modify their links).  The damage would
not be restricted to just lost orders -- the damage to company
reputation would be formidible, as well.  A one day outage would be
major news, and terrible publicity.  A week long outage would have a
serious impact on the companies future.  A month-long outage could
destroy the company. 

Removing a domain name is just deleting a few entries in a zone file,
and such an outage could be instituted in a moment by the .com
registry.

There aren't any close analogs to this situation in realspace.  A
landlord can't boot you out of your realspace storefront in a
millisecond through an editing glitch -- eviction takes significant
due process.  Real estate law is complex, with protections for both
renters and buyers; transactions are heavily regulated, through
zoning laws, laws concerning mineral and rights, tenants rights, and
other things; if you are concerned about being locked-in by a
landlord, you can always buy outright.  None of this exists with
domain names. 

Moreover, transactions with real property involve large sums of
money, and the cost of moving is comparable in cost to the
transactions involved.  This is not the case for a domain name -- the
cost of a domain name is miniscule, but the value to the registrant
is potentially enormous. 


                           The FTC paper

Given these realities, then, the FTC's bland analysis of lock-in
leaves a lot to be desired:

  FTC: "It would appear plausible that the absence of domain name
  portability across registries could impose a switching cost on
  users who change registries.  For example, if a firm must invest
  substantial resources to familiarize consumers with its web-site
  name (e.g., "brandname.biz"), the cost of switching to a new site
  (e.g., "brandname.store") would consist of the incremental
  investment that it would have to make to inform consumers of the
  new name, plus any lost profits from forgone sales (because some
  consumers never learn the new site)."

Implicitly, the FTC makes the same erroneous assumption that Mr Feld 
and others have made, that you always have the option of keeping the 
old domain while transitioning to the new. (*)  That is, the vendor 
has available the ability to shut down your business totally.

The FTC also failed to consider the vast disparity between value and
cost -- maintaining a domain name record is, for practical purposes,
costs very close to zero, while the value to the company may be in
the millions of dollars.  That leaves a lot of room for price
increases that are well below the threshold of visibility. 
Concretely, the cost of a domain name could be increased by a factor
of ten, and still not equal the cost of having a legal professional
do even a cursory review of the situation.

Even in the friendly case, where there is plenty of time to make a
move, a company is not going to want to give up the old name for a
long time.  (**) Given 3 million web pages with links to amazon.com,
what is the cost to Amazon to switch to amazon.biz *and* turn off
amazon.com? At $35/year, or even $350/year, there is no reason to
ever turn off amazon.com, and no matter when you do it, it instantly 
cuts off some portion of your revenue stream.

In fact it appears that the economists involved in th FTC paper were
hurried, were not familiar with the Internet, and simply quoted
general studies on the subject of lock-in.  They devoted very little
time indeed to consideration of the concrete situation with domain
names.  (***)

It should be noted that the "lock-in" we have been describing is a
structural feature of the market, not something that a company does. 
NSI does not cause lock-in -- lock-in is cause by the fact that
domain names are one-way links embedded in data that is outside the
control of the owner of the domain name.  That is, lock-in is an
intrinsic and unavoidable feature of the present usage of domain
names.  Competition between registries or registrars has no effect on
lock-in; lock-in in the domain name system is technically and
socially determined, not economically determined.  (That is not to
say, of course, that this lock-in has no economic effects.)

This kind of lock-in is somewhat different than what the FTC
describes: "The phenomenon arises when buyers must make
relationship-specific investments in order to do business with
particular suppliers."  It may be that the FTC economists were asked
to produce a paper on "lock-in in the domain name system", and, 
having only a cursory familiarity with the domain name system, 
simply wrote a paper about lock-in as a general phenomenon.  
The USG has in general not exhibited a lot of sophistication in 
their understanding of some of these complexities.



                       ================

(*) The suggestion has also been made that there is a technical fix 
by using http redirects.  This suggestion fails for the same reason 
-- if you don't control the domain, you can't cause a redirect to be 
there.  But it also fails because it doesn't address other uses of 
domain names, such as email...

(**) I notice with some nostalgia that morningstar.com has moved to
morningstar financial services -- morningstar.com was the domain name
of a network company that did one of the first implementations of the
PPP protocol, and manufactured routers based on that technology. 
Morningstar was bought out long ago, but for many years thereafter
you could still find the documentation for morningstar routers on
that web site.

(***) Incidentally, the FTC report also addressed the issue of
non-profit vs for-profit registries, and committed the same sin --
they failed to analyze the reality of the "non-profit" scenarios
proposed, and, once again punted to general academic studies on the
subject.  In particular, they failed to consider the dynamics that
obtain when registrars have significant control over registry policy. 
The FTC paper also fails to consider the following: 1) most of the
improvements claimed for for profit registries also come with a
scheme that allows for profit registry *operators*, and non-profit
registries, with all other functions delegated to competitive
registrars; 2) the registry function in the "non-profit" models is a
fairly narrow one; 3) "competitive registries" is a meaningless term
unless you allow for the possibility of registry failure.  Registry
failure is not conducive to the "stability of the Internet", the
primary constraint on ICANN (both from the White Paper, and from
later expressions of the ICANN board.)

-- 
Kent Crispin                               "Do good, and you'll be
kent@songbird.com                           lonesome." -- Mark Twain