After eight months, similarity review creates only TWO new gTLD contention sets
ICANN has finally delivered its String Similarity Panel’s review of all 1,930 original new gTLD applications, finding that only four applied-for strings are confusingly similar to others.
Two new contention sets have been created:
- .hotels (Booking.com B.V.) and .hoteis (Despegar Online SRL)
- .unicorn (Unicorn a.s.) and .unicom (China United Network Communications Corporation Limited)
Only one applicant in each contention set will survive; the strings may go to auction.
The list is bafflingly short given that the panel’s review was originally due in October and has been delayed several times since.
ICANN last month said it was forcing the panel to address “process” issues and heavily suggested that it was trying to make sure the review process was legally defensible, leading to speculation that the list was going to be much longer than expected.
But the panel and ICANN appear to have taken a super-strict approach to finding similarity instead.
The string similarity panel was tasked with deciding whether each applied-for string “so nearly resembles another visually that it is likely to deceive or cause confusion”.
I think the four strings included in its final report pretty conclusively pass that test.
Even the controversial Sword tool – the software algorithm ICANN commissioned to compare string similarity on objective grounds — agrees, scoring them very highly.
According to Sword, .unicorn and .unicom are 94% similar and .hotels. v .hoteis have a score of 99%. For comparison, .hotel versus .hotels produces a score of 81%.
It all seems nice and logical and uncontroversial. But.
All four affected applicants are applying for single-registrant gTLDs, in which the registry owns all of the second-level domains, drastically reducing the chance of abuse.
One of the reasons ICANN doesn’t want to create too-similar TLDs is that they could be exploited by phishers and other bad guys to rip off internet users, or worse.
But with single-registrant TLDs, that’s not really as big of an issue.
It will be interesting to see the affected applicants’ response to these latest findings. Expect complaints.
The results of the review will be a huge relief to most other applicants, which have been wondering since the list of gTLD applications was released last June what the final contention sets would look like.
The high standard that appears to have been used to find similarity may set some interesting precedents.
For example, we now know that plurals are fair game: if Donuts’ application for .dentist is approved in this round, there’s nothing stopping somebody else applying for .dentists in a future application round.
Also, brands with very short acronyms have little to fear from from being found too similar to other existing acronym dot-brands. The applicants for .ged and .gea were among those to express concern about this.
But the question of confusing similarity is not yet completely settled. There’s a second phase.
Applicants and existing TLD registries now have about a week to prepare and submit formal String Confusion Objections, kicking off an arbitration process that could create more precedent for future rounds.
Unlike the just-published visual similarity review, SCOs can take issue with similar meanings and sounds too.
Losing an SCO means your application is dumped into a contention set with the winner; if you lose against an existing TLD operator, your bid is scrapped entirely.
If you find this post or this blog useful or interestjng, please support Domain Incite, the independent source of news, analysis and opinion for the domain name industry and ICANN community.
Um. So that’s like 15 seconds work, right? The first is (I guess) a typo.
And this took how many months?
.hoteis is Portuguese for .hotels, I believe.
Indeed is. But the .hoteis applicant also applied to .hotel (same word in english, portuguese and spanish) and to .hoteles (spanish for hotels).
.hotel is a contention nightmare with 7 applicants total, but .hoteis only compete with .hotels. It seems logical to get some money from the .hotel private auction and outbid .hotels (either in private auction or ICANN auction) ending up with .hoteis (portuguese) and .hoteles (spanish).
I wouldn’t confuse the brevity of the list with the amount of work it took ICANN to reach these conclusions. The result is spot on IMO. Imagine a namespace where singular’s and plurals can’t happily co-exist.. That was the alternative here. However it is they drew these conclusions, kudos to ICANN for cementing the viability and long-term success of NewTLDS. We are moving up a level – and it will happen in most of our lifetimes.
I totally agree with Frank’s comments. By including another 50 that we identified, as in .car ( Google) with .cars (Donuts,Uniregistry,DERCarsLLC) would only clog up the existing 230 +4 contention set resolutions further.This way is fairer and will ultimately lead to further competition and choice for consumers. Finally all 234 strings / 758 applications can move forward and start working out their contention set valuation/pricing/ JVs/funding/ private or ICANN auction strategies. The fun begins. Deep pockets will win.
Interestingly the “visually similar applied-for gTLD strings” is only mentioned in the IDN section of the guidebook (where it make sense). I would argue the the “create a probability of user confusion” is more for the ASCII gTLDs. I think ICANN has made it the easy way, mixed both things together and that’s the wired outcome we know.
“All four affected applicants are applying for single-registrant gTLDs, in which the registry owns all of the second-level domains, drastically reducing the chance of abuse.”
So how do you know that they will be single registrant TLDs always? What if they decide to change their minds at a later date? Will the risk of abuse go back up?
That’s what they all say in their applications, according to the quite marvelous DI PRO database.
I personally think that ICANNs decision will have impact to domain name disputes worldwide and the sword tool may become a standard tool in discussions about trademark similarity on the Internet.
If the tool ever becomes trustworthy. ICANN makes no claims to the accuracy of the tool. They only say it should be used as a guide
SWORD only looks at visual similarity, and most of the UDRP cases I saw were about composed phrases of exact match words, like bmwsoccer.de
Objection Time
Someone should apply for ‘.coms’
Given the IRS Form 990 that was published, we know that the String Similarity Evaluations cost the public a whopping $6 million:
http://www.icann.org/en/about/financials/tax/us/fy-2013-form-990-en.pdf
(see page 64, 2nd line) We also know that 2 contractors were selected to handle string similarity:
http://newgtlds.icann.org/en/blog/preparing-evaluators-22nov11-en
namely InterConnect Communications & University College London.
Does $6 million seem excessive to anyone besides myself??!!??
1930 * (1930-1) string confusion possibilities would give 1 dollar and 60 cents for each analysis. The numbers are not that high due to exact matches and to combinations with too low SWORD index, but the numbers don’t sound excessive to me. I do think though that they agreed on a rate before knowing that there would be so many applications, and that re-work costs are probably also a factor.