What's your proposed alternative? That generic category names should be disallowed within a TLD? Or that some public or community-driven version of a for-profit service should control the generic names?
Maybe names that have obvious “generic” squatter value—i.e. names that (at the time of their attempted purchase) describe entire verticals, rather than uniquely identifying a business—should be put up for auction rather than being first-come-first-served.
(That would also include names that describe a vertical that someone tried to trademark into being the name of a business anyway: pets.com and the like. If that trademark isn’t already well known, then this is just an attempt to end-run the auction system, rather than a valid claim.)
Given that TLDs can decide their own allocation policies, I’m surprised none of them have tried this yet.
Well the .dev domains are going up in a "dutch auction" style.
The price starts off at an insane $11,500 per domain, and then slowly over the course of 2 weeks goes down to $12 per domain, first come first serve.
It's a pretty good solution in my opinion to the problem of how to "fairly" distribute domain names.
I want to pick up a few, but I'm going to wait until they are down to the lowest tier as i'm pretty sure nobody is going to pay even $100 for them. And if they do, well then they probably want it more than I do!
>should be put up for auction rather than being first-come-first-served.
Have you by chance heard of the Coase Theorem in economics? Which states that given that property rights are well defined and transaction costs relatively low, initial distribution of the property rights will not affect the efficiency of the final outcome reached. In other words, if a given company is given the rights to something and some other company would value that thing more, we do not have a problem, they will negotiate to the efficient outcome on their own by trade.
Why? Even if it's "cute", it serves no other purpose than racking up prices and keeping buyers second-guessing themselves. Just run a normal Dutch or a second-price auction.
If your question is why not run a standard auction, the answer is that operational overhead to do that is much higher than simply having a given price for creations on a given day, and there'd thus be fewer registrars participating. You need a whole eBay-like experience to run a standard auction; it's a non-trivial amount of work just to support one launch.
I stand corrected on the name. The operational overhead might be reasonable excuse, but I don't think a second-price auction would necessitate that much effort. You could get by with just a form and running a query on it.
It's a LOT of effort. And said effort is required not just at the registry level, but also at the dozens of registrars that support EAP.
This isn't a theoretical concern. We actually did use standard auctions for our first three TLD launches. They were operational nightmares. The Dutch auctions we're using now are much more light-weight, have greater participation, and are now the preferred way of doing launches across the entire industry. We're not even the ones who started using Dutch auctions for TLDs.
You'll just have to trust the word of someone who's been doing this stuff for years that the alternatives really are worse.
The thing is that DNS infrastructure (which registrars basically rely directly on) is very OLAP oriented. bind(8) and DNS daemons like it are DBMSes in some sense, with replication and serving highly-concurrent reads being their primary focuses, and inserts/updates/deletes taking about fifth place compared to those needs. And registries (and sometimes registrars as well) build their entire infrastructure to use the DNS-daemon “store” as the canonical store, rather than having it be a secondary system synchronized into from an online OLTP DBMS. So this writes-are-expensive paradigm creeps into the entire DNS infrastructure, including things seemingly far away from the core, like the registrars’ control panels.
A good comparison is blockchains, which are also highly-replicated, cheap-reads expensive-writes DBMSes. CryptoKitties chose Dutch auctions to sell their kitties for a similar reason to that of the .dev registry: since putting state into their “DBMS” is so expensive, they wanted as stateless a system as possible. (In their system, even the current price is just statelessly computed by comparing the client’s time-at-bid to the chain’s recorded start-of-auction time. The only writes needed are to start the auction, to place the winning bid, or to cancel the auction.)
Probably the former as it doesn’t require as much aggrandizement.
I wonder whether a different approach would work. Instead of basing TLD name spacing on nebulous terms (com,net,org,io,edu,etc) create a naming system that better reflects the real world. The abstracted terms lead to clashing. So addressing actually reflects physical addressing. There’s only ever going to be one company at Apple’s corporate address. Not sure how it’d work.
Edit-the tech to do this via gps even exists today.
The irrelevance of physical location is one of the advantages of the Internet; adding that to addresses just to provide an unique key seems absurd. Would a company have to change all their addresses if they move their headquarters? What if the country's street system changes? And what about individuals?
I don't see a need to change the current system, but I'd rather move to random strings than tying them to some irrelevant and mutable piece of data.
We could just give everyone who wants a domain a UUID.
My site could be 'uuid://09953e01-8164-4e40-8504-ed5507b50b03'.
This seems to solve all of your complaints. No clashing, reflects the real world (names are all meaningless is reality, and physical locations are nebulous).
> There’s only ever going to be one company at Apple’s corporate address.
On the other hand, there are corporate addresses that have many companies, and many domains, and many distinct technical departments with different needs.
What you're missing here is that domain names are meant to be easy to remember. That is their only goal.
They were created so people didn't have to memorize ipv4 addresses. Proposing that we use corporate addresses or gps coordinates is absolutely unrealistic and against the spirit of the domain name system in the first place.
> They were created so people didn't have to memorize ipv4 addresses
New problem: memorising a UUID is even harder than memorising an IP address. Someone will eventually invent a DNS for the UUID-space, and we'll have this problem all over again.
If anyone actually wants to test this: IPv6 is already a modified version of this UUID plan. The designers were even kind enough in designing IPv6 to provide some nice memory shortcuts over UUID in the IPv6 address style, such as :: to fill runs of zeroes.
It might~ be fun to see how much of the web you can explore just memorizing IPv6 addresses rather than using DNS.
I know we're both just joking, but unfortunately due to SNI headers and load balancers, memorizing ipv6 addresses doesn't necessarily get you the right site.
The DNS system has also managed to make ip addresses less stable over time.
