The two of us agree on some things, and disagree on others. We agree that:
The best thing for the internet is to release the next batch of
gTLDs all at once, and to have that batch be about 50 new gTLDs. That
way, no one gets much advantage of having a new TLD other than the
semantics of the TLD's name.
Anyone even considered being awarded a gTLD needs to prove in
advance that they have the technical ability and skills to run a
gTLD. This means existing DNS nameservers in geographically and
topologically diverse areas, lots of bandwidth to those nameservers,
and the ability to maintain a large number of queries on each one.
Fortunately, these numbers can be agreed to ahead of time, and simple
tests can be created to test the applicants.
This is like the qualification process for a CLEC (competitive
local phone company) in the US.
In many cases, we expect applicants would simply contract with someone
that already runs other TLDs to provide the technical infrastructure.
ICANN should not profit from releasing the new gTLDs.
If ICANN thinks it needs a bigger budget, it should go through a budget
process to justify that budget, not just get windfall money from
the gTLD process.
The winners of the new gTLDs get no exclusivity over any other
names issued by ICANN. That is, the first round might include
.rugs and .carpets, or the first round might include
.tooth and the next round might include .teeth; the
winner of .rugs cannot prevent .carpets from being
Generic (non-country) TLDs have failed at being directories, and
that failure is getting worse over time. Users will always prefer to
use search engines to find which sites relate to a particular topic
than to assume that only domain names with a TLD that is semantically
linked to that topic are of interest.
There are probably only two plausible routes to a successful TLD:
.com clones and certification. The clones are just like .com, only less
crowded; we already have two of those, and a few more could probably be
useful. The other route is certification. The reason that .edu is a
success while .coop and .museum aren't is that people care whether
something is an actual degree granting institution, while few people
care about "real" museums and "real" co-ops. ("What a fool I was, they
said they were a co-op but really they were only a producer's
collaborative.") For a certified TLD to be useful, it has to cover
an area that is relevant to many Internet users and be managed by
an organization that will ensure that only bona fide SLDs are issued.
Some of the most useful domains like .edu are not
particularly large or lucrative. None of the proposed schemes (including
our own) are likely to find them. If there is a place for creativity to
be focused, it should be on figuring out which new TLDs are the most
useful to typical Internet users and make sure those TLDs exist and are
The two of us disagree on the best way to make these bunches of 50
John sees two routes to selecting TLDs. For TLDs intended to make
money, the best approach is an auction, with the N highest bids getting
to pick their N favorite domain strings, and the money given away to a
suitable worthy cause, not ICANN. Other people have made more detailed
proposals to deal with the obvious trademark issues, e.g., only IBM can
pick .ibm but they still need a winning bid to do so. As Paul
notes below, ICANN's beauty contest has picked losers, and a lottery
tends to turn into auctions where the lottery winners keep the auction
proceeds. Possible approaches include a separate lottery for five or ten
names for which only non-profits can apply, giving virtuous bidders
funny money they can use in the auction, as was tried in the PCS
frequency auctions in the US. John doesn't have any great confidence
that these will work, but if the auction process can be made simple and
predictable enough, it should be possible to try one approach this year,
another next year, and so on until one turns out to work.
Paul believes that there is no way to predict which TLDs might be
"best". The track record so far is abysmal.
Having an auction might get people to think harder about which gTLDs
would work best, but it is completely unclear who should profit from the
auction. Instead, a lottery based on the desires of the organizations
who qualify to be gTLD owners could be designed to get a wide variety of
TLDs, with some organizations becoming big winners and the rest having
ones that don't cost much to run. A lottery would prevent ICANN from
making unnecessary money on the system, and would open the market to
many companies who might otherwise be locked out.
Both of us agree that once the 50 gTLDs are assigned, there will be a
lot of buying and selling of assets, regardless of what the rules for
the auction or lottery say. Just live with it; that is how big business
works. But the values of the new gTLDs will be much lower than might be
expected because there are so many of them, with maybe another 50 or 100
a year later.
A little over 42 years ago, in Martin Luther King Jr.'s acceptance speech for the Nobel peace prize, he said:
. . .
After contemplation, I conclude that this award which I receive on behalf of that movement is a profound recognition that nonviolence is the answer to the crucial political and moral question of our time - - the need for man to overcome oppression and violence without resorting to violence and oppression.
. . .
I refuse to accept despair as the final response to the ambiguities of history. I refuse to accept the idea that the "isness" of man's present nature makes him morally incapable of reaching up for the eternal "oughtness" that forever confronts him.
. . .
I believe that unarmed truth and unconditional love will have the final word in reality. This is why right temporarily defeated is stronger than evil triumphant.
Apple shipped its first Intel-based Mac today, and will ship another one next month. They're much faster than the current breed of Macs, and have a few extra nice features.
Unfortunately, there was no Intel-based virtualization software announced or even hinted at. When Apple announced the coming Intel-based Macs last June, I wrote, twice, about why having software that let a user run Windows on their Mac as a sub-process under OS X would make Apple lots of money.
There are lots of corporate users who, as individuals, have bought Apples (correct) line that Macs are easier to use, more fun to use, and much more reliable than Windows-based PCs. But their IS folks won't buy them a Mac because the Mac can't run the custom in-house programs, or the Windows-only VPN client, or the something else that IS has decided makes you part of the company.
Reasonable virtualization (and the Mac version of VirtualPC is not reasonable) fixes that. It gives the already-sold employee another tool to wheedle with. Guessing numbers of potential users that would turn into sales is impossible, of course, but those will be long-term repeat customers as well as first-movers in companies.
For all I know, Apple may already be on top of this and is funding the work, but it's just not ready now. Or, they may have already walked away from the market, but some smart company like EMC is doing it without Apple. For example, if VMware player (which is free) ran under OS X, someone could sell bundled (legitimate) Windows XP images. If Microsoft makes Virtual PC for the Mac run on Intel-based OS X, they could sell it all without cannibalizing any sales.
We don't know yet. All we know is that there are some Mac sales not being made because this market segment is not yet possible.
It's been a long time since I chose to use a managed service instead of being my own sysadmin. I have moved my self-hosted blog to TypePad, and have (I believe) made all the old links redirect correctly. There may be some glitches, and I'll certainly experiment with the layout here.