Sales/Auction models for the permanent registrar


#1

The permanent registrar will require some form of mechanism to allocate new domain names to registrants. The current interim registrar operates off a vickery auction system, with four operations: (1) Start auction, (2) Bid, (3) Reveal, and (4) Finalise. Bidding and revealing are time-sensitive, and a failure to reveal can lead to loss of funds.

One proposed improvement to this system is the “rolling auction” mechanism I described at DevCon 3. In this system, no start auction action is required, and any name can be bid on at any time. Bids are then revealed in the subsequent time window, with the winner and the price being decided using the same mechanism as previously. In the event that someone bids in the same interval that someone else reveals, the late bidder is refunded their whole bid and does not take part in the auction.

This has the advantage of reducing the number of actions required from four to three, and simplifying the schedule for reveals, but reveals remain time-sensitive and the possibility of lost funds remains.

I’d also like to re-raise for consideration my proposal at the first ENS workshop, that auctions should be used when new names are made available - such as when names shorter than 7 characters are made available - but that for day-to-day registration we should revert to a simple fixed fee (or no fee, rent-only) model. Experience has shown that the auction process is a significant barrier to entry, because of the number of interactions required, the time sensitivity and risk of revealing bids, and because of the mandatory time delay before obtaining a domain.

I believe that the evidence post-launch demonstrates that the base rate of contention - multiple bids on the same name - fell substantially when the soft launch concluded, and isn’t high enough to justify mandatory auctions on all names. Those that still attract multiple bids are likely to be names that generate sudden attention such as news events and memes, and I don’t believe the value of preserving ‘fair’ allocation of those names outweighs the usability expense of doing so.

Thoughts?


#2

When auctioning names with the current system the issues appear to be around the complexity of the Vikrey auction regarding the number of transactions that need to be made, and the time sensitive nature of revealing bids. Is there any downside to changing the auction such that a user who has submitted a bid but forgets to reveal it can unseal it at any time in the future and obtain their 99.5% return rather than be punished so severely for forgetting to reveal? This would allow users who forget to reveal during the auction period to at least receive their money back.

Having two different systems to purchase ENS domains (an auction and a buy-it-now) would be confusing, especially if the two systems run side-by-side for different domains. What if prior to the permanent registrar going live a pre-auction for the to-be-released domains took place? It would work much as the current auction system does, but the end result would be a reservation for the given domain rather than the domain itself. At the time that the permanent registrar goes live the reservations would be translated in to domain ownership, and all future domain purchases could be on the buy-it-now principle? I think that this would be clearer for users and would have the additional benefit that the permanent registrar would be simpler code-wise.


#3

The problem here is that it enables an extortion attack: I bid at a series of different prices on a name, then when you reveal your bid, I threaten to reveal one just below yours to force you to pay as much as possible.

That seems like an excellent idea.


#4

It seems that most of the users complaining of the complexity of the system is based on their lack of time management. If a domain is important for an aspect of someone’s business or true desire to acquire, there should be some hurdle in acquiring. I actually, loved the Vickrey auction.

The benefit for the system is two-fold.

First, the acquiring party shows some basic technical ability to understand, and manage time, and at least some commitment to a potentially valuable asset. Secondly, it will reduce squatting (unless someone turns it into a full time gig). With the addition of ENSnow there will be ample opportunity to acquire a subdomain .eth with less technical skill. Having side-by-side systems or changing the existing auction to reduce a step does not seem necessary and may lead to further “confusion”.


#5

If an additional system is implemented, maybe the initiator of the auction needs to be the first to reveal, and if within a time time period fails to do, the entire contract would expire with a % refund. That would require all secondary bidders to have to wait till the first reveal and then decide (or not) to reveal theirs. Or the secondary bidder can just wait out an additional secondary reveal period, and then the entire auction could expire. Then it can all be restarted by the forgetful initiator, or someone new.


#6

Do we gain anything by excluding people that ‘fail’ this test? Erecting artificial barriers to participation seems like a bad idea to me.

I also want to discourage the view that names are ‘assets’, personally. They’re a way for humans to find resources, not property.


#7

I agree the .eth domains should not be viewed as assets, just locator tools to resources. I meant “asset” in the context of “a useful or valuable thing” not as “property owned by a person or company, regarded as having value and available to meet debts”, it is an import distinction you have pointed out.

Additionally, you probably won’t gain anything from keeping the auctions as is. My point was that the Vickrey auction is not too complicated to figure out, and people love to complain even if the complaints are outliers. Usually people who are happy just go on their merry way because it all went as planned. It’s the squeaky wheel that get the oil. I understand.


#8

Due to the low contention of ENS names at current (squatters notwithstanding) people are used to being able to place the minimum bid and expect to win, which is a problem. That gives them a bit of bad learning and when the squatters do take their name they are upset. It also, incidentally, gives the squatters a free ride compared to if bidders bid correctly under a Vickrey auction system.

I think a good approach is to keep the current Vickrey auction with the tweak that Nick mentioned to cut down on the number of transactions created, and to write an article that clearly explains to those who are using the reservation system for the yet-to-be-released names how they should bid (i.e. bid maximum you would pay rather than minimum you believe you can get away with). The article can be linked from the main ENS bidding sites so that the how-it-works is as clear and visible as possible.

In addition to this, finding a way to help people remember to reveal bids, or to reveal bids on others’ behalf, would be handy. Perhaps working with mycrypto/mew to give ENS bidders ICS files with calendar reminders to reveal their bid. Alternatively, something discussed back at the ENS workshop was some sort of proxy bid system that would allow one user to reveal on behalf of another but I haven’t thought that one through to see if it’s realistic.


#9

I think we should expect the current situation - very low contention except for opportunists - to be the norm post-launch. Do you agree? If so, is there any point in still having auctions?


#10

Sorry yes, I’m thinking of the pre-registration of the <7 character names with the auction. I’d be more than happy to see the permanent registrar go out with a fixed-price system.


#11

What is the point of having auctions?

  1. reduce squatting?
  2. allow market to initially value domains?
  3. any others?

#12

There is a laissez-faire perspective, socialist perspective and a tightly governed perspective.

This may take some drawn-out critical thinking, analysis and group participation; but here it goes: From a laissez-faire perspective, all names should become available on a “first come first serve” basis, priced for a nominal fee and let things take their own course, without interfering.

From a more restrictive perspective, everyone should have to “prove” they have a bonafide interest in the .eth they seek and then all those who have a real “connection” to the mark (not speculation) could bid against each other (kind of like the current Vickrey auction system but more restrictive).

Then after a vetting period, all vetted participants would bid against each other for the rights to the .eth use as a beneficial tool for their Ethereum project. It’s kind of like how US trademarks are issued. The registrant picks a category of use like automobiles, hotels, or clothing, and then attests to its use in commerce as 1) “Actual Use in Commerce”, or a 2) “Bonafide Intent to Use in Commerce”. For ENS there could be third category where the registering party has to show some connection to the name as a bonfide real name or pseudonym.

This creates layers of complexity that we may have to address at some point. For example, Ford is a motor car company. Ford is also a completely separate worldwide modeling agency. Ford is also a common last name. Who should have the rights to Ford.eth (i know its less that 7 letters, its just an example).

Now under US law if Jane Ford (an individual), decides to start using the mark Ford on her new jewelry line and registers FordThings.com FordThings.eth everything might be fine. She sells her jewelry and makes some disclaimers along the way that her products are NOT produced in relationship to any Ford modeling agency or Ford motor car company. But it’s not so simple because people get all possessive about stuff and in many cases rightfully so. Does Apple Computer really “own” the word Apple on everything? even an apple? Does Apple Computer own Apple’s apples?. Would (should) Ford motorcars, Ford Modeling, and Jimmy Ford (an individual) have superior rights to the name or mark, over Jane Ford’s new jewelry line?

It’s super complicated that results in millions of dollars in litigation every year in fighting for “rights”. If the Ethereum Name Service keyholders https://ens.domains/index.html are expected to arbitrate these challenges, God bless them.

I believe it makes sense to be somewhere closer to a laissez-faire or socialist perspective public, collective or cooperative ownership, or to citizen ownership of equity. Like taking certain popular domains and putting them into the ENSnow system and shairing all the proceeds with the ENS user group (https://gitter.im/ethereum/go-ethereum/name-registry).

In any case, the Anticybersquatting Consumer Protection Act (ACPA), 15 U.S.C. § 1125(d), is a U.S. law enacted in 1999 that established a cause of action for registering, trafficking in, or using a domain name confusingly similar to, or dilutive of, a trademark or personal name should be used a a general framework for any .eth accounts under consideration for the blacklist.


#13

Where multiple people want to register the same name, it makes sense to try and find a ‘fair’ way of determining who gets it. The two widespread systems are “first in first served” and “auction”; the latter seems fairer to many (myself included) because it allocates the name to the person with the most financial interest in it, and finds a market price for it. Allocating it to the first person to grab it encourages squatting and other predatory practices taking advantage of the difference between the cost to them and the market value.


#14

Between Ford Motor Company and Ford Modeling Agency bidding for FordThings.eth this may be true, but average citizen Jane Ford and her little jewelry line will in no way every be able to compete. She is in essence eliminated from the opportunity. I have no opinion if that’s a good or bad thing.

Of course they should all just settle for FordCars.eth FordModels.eth and FordJewelry.eth in each of their own “spaces”. That leads back to a first come first served system. Conventional wisdom tells us, shorter is easier to remember. Maybe Apple.eth is easier/better than AppleComputer.eth, who knows?

The alternative of the auction leads to certain names need a bidding war for the shortest easiest to remember Ford.eth , (just a short example) and maybe rightfully so. Apple.eth i assume is worth a bidding war. Except for Apple, the nice lady who sells apples, who will have to settle for ApplesApples.eth


#15

I agree that while it was a good experiment to have a full Vickrey Auction, the complexity really shined people away from registering top level domains and the mania led to very high prices. I would like to have one click auctions too, but I’d want a single system for everything. Why not have a simple linear bid?

  1. Bob bids whatever amount he wants, on the open, but on a hashed name.
  2. If in 12 hours, nobody else makes another bid, bob keeps his domain. When he finalizes it, he will be refunded his full bid minus the minimum price
  3. During the first 12 hours, anyone can bid, as long as they increase at least 20% the bid price from Bob. When that happens, a new 12/24h deadline is set
  4. If the bidding war lasts more than a predetermined amount of time the minimum bid price goes to 50% or 100% over the last bid, to force the bidding war to stop
  5. When the deadline is over, the domain belongs to the last bidder, no further action necessary
  6. when the auction is finalized (optional) then the owner gets a return on his last bid - the price of the last bid. In that sense it works like a second price auction, where your current bid is just a maximum

This enables sales to be made with a single transaction. For most, uncontested auctions, it only adds the burden of the 12 hour waiting time. For the rest, it allows a bidding process. It would be the same for newly released names, unlocked names or new names.

This is the model that we were trying on the ens domain resale site.


#16

I like the fact that that proposal has no situations in which someone can lose their funds without getting a domain, but I still think it’s too complex for day-to-day use:

  • It requires a minimum of 2 transactions (bid and finalize).
  • It requires a minimum delay of 12 hours (and 12 hours seems low for other people to spot the auction and bid).
  • For popular domains, it may lead to more onchain transactions than the Vickery auction.
  • It still leaves opportunities for people to take advantage of others’ irrationality by bidding on names and ransoming them.

But I still think the most compelling argument against such as system is the very low rate of contested domains post-launch. When significantly fewer than 5% of domains have more than 1 bid, the additional complexity of an auction system seems hard to justify.


#17
  • Finalizing is optional, you need it to setup ownership but you are already owner. We can even make it so that anyone can finalize domains for others.
  • Well we could also have 24h windows so it works no matter your timezone. It would take away the chance of doing a instant joke by creating a domain for something that is breaking news, but I don’t see that as that bad
  • Is the number of on-chain txs a problem? We could think on how to do it off chain, but it adds complexity
  • This is a point I agree, I would tend to think that sealed bids would lead to more thoughtful pricing, while running bid wars would lead to more emotional impulsive bidding. But some of the prices on the Vickrey auction were quite irrational anyway!

I understand the point of the low level of contested auctions but we will have to have some sort of auction anyway, when we release a new batch of names. So why add complexity by having the new batch play by different rules than the others? In this model, if 95% goes uncontested, then they all took a single tx anyway, all we did was ask them to wait 24h to claim it.


#18

Sure; it’s still an additional transaction for each registration, regardless of who executes it.

I quite like @jgm’s suggestion - run an auction for short names before the permanent registrar is activated, then migrate everything over to that when it is. The permanent registrar doesn’t need any auction mechanism at all included.


#19

Over engineering this system will not lead to a smooth widespread adoption. Without addressing the individual “value” of select X.eth accounts, the auction system will be the main barrier standing in the way of widespread adoption, which I think is part of the goals of ENS in the first place. The more I think about it, the “first come first served” system is the one that will allow the most users to flood into the environment. There are already laws set up to protect people against co-opting a protected name/mark (ACPA passed in 1999). I know it’s not a sexy developed protocol, but over engineering can slow innovation and user adoption. If the most well know names are identified and embargoed to the best of the ENS group’s skill, this can at least divert some of the squatters actions.

“Metaphorically speaking, to open a can of worms is to examine or attempt to solve some problem, only to inadvertently complicate it and create even more trouble. Literally speaking, opening a can of worms, as most fishermen can attest, can also mean more trouble than you bargained for.”


#20

If the most well know names are identified and embargoed to the best of the ENS group’s skill, this can at least divert some of the squatters actions.

We did not single out any name, not even ethereum.eth and I hope we can keep it that way. A hand picked list is the can of worms you’re describing because there’s always some name left out and at some point it simply becomes a central system where we decide all owners.

run an auction for short names before the permanent registrar is activated, then migrate everything over to that when it is. The permanent registrar doesn’t need any auction mechanism at all included.

I’m not against having a first come first served system for names that have been available for a long time and nobody wanted. Maybe having auctions for newly released and then it’s all fcfs might be good enough.

Another interesting model that can be applied either for new names or recently released would be a reverse dutch auction: all names start at a 100 million ether price, instant purchase, and it goes down until it reaches 1 Wei (or some other minimum) after six months (reducing the price about 10x each week takes us to 100M ether to 1 Wei in 26 weeks). This combines the slow release we had before, with a single click auction. This could either be done only on newly available names or even with released names, in the second option, the names would start at previous price.

A problem is that if all names start at the same time, then it can be true that names will be unreasonably high for weeks until they become super cheap in a matter of a few days, at which point most activity will be concentrated. Maybe the starting price could be random, like starting dates were.