String confusion in disarray as Demand’s .cam loses against Verisign’s .com

Demand Media is demanding an ICANN review of its objections policy, after its applied-for new gTLD .cam was beaten in a String Confusion Objection by .com registry Verisign.
A International Centre for Dispute Resolution panelist has ruled (pdf) that .cam and .com are too confusingly similar to coexist, meaning Demand’s bid for .cam must be rejected by ICANN.
But the ruling by Urs Laeuchli conflicts with two other ICDR panel decisions on .cam, which both found that the string is NOT confusingly similar to .com and therefore can be delegated.
So while Demand’s .cam bid, under a strict reading of the rules, is now supposed to be rejected, applications for identical strings filed by AC Webhosting and dotAgency can go ahead.
ICANN has been thrown a curve ball it is not yet fully prepared to deal with.
As Akram Atallah, president of ICANN’s Generic Domains Division, told DI last week, it’s possible that the policy or the implementation of that policy may need to be revisited by ICANN and the community.
United TLD, the Demand Media subsidiary that applied for .cam, is now calling for precisely that, with vice president of business and legal affairs Statton Hammock writing today:

String confusion objections are meant to be applicant agnostic and have nothing to do with the registration or use of the new gTLD. What matters in string confusion objections is whether a string is visually, aurally or, according to ICANN’s Applicant Guidebook, otherwise “so nearly resembles another that it is likely to deceive or cause confusion.” Individuals may disagree on whether .CAM and .COM are similarly confusing, but there can be no mistake that United TLD’s .CAM string, AC Webhosting’s .CAM string, and dotAgency Limited’s .CAM string are all identical. Either all three applications should move forward or none should move forward.

The .cam cases are not alone in presenting ICANN with SCO problems.
Last week, Donuts’ bid for .pets was ruled confusingly similar to Google’s .pet, despite previous ICDR cases finding that plurals and singulars are not too confusing to coexist.
Where the .cam panelists disagreed
While there were three .cam cases, two of them were decided by the same panelist. It seems that both panelists were provided with very similar sets of evidence in all three cases.
It’s relevant to note that neither panelist — unlike some of their colleagues in other cases — thought it was appropriate to apply trademark law such as the DuPont factors in their decisions.
They did, however, consider the expected use cases of .cam.
All three applicants take .cam as short for “webcam” or “camera” and would target registrants interested in those fields (a lot of the use will likely be pornographic — AC Webconnecting is a porn firm after all).
But all three applicants also want to run “open” gTLDs, with no registration restrictions.
ICDR panelist Murray Smith was in charge of both the AC Webconnecting and dotAgency cases. He addressed expected usage explicitly in dotAgency, and explained why:

It is not just the visual, phonetic and conceptual similarity between the words that must be taken into account. In my view the greater emphasis should be focused on the use of the disputed extensions in the context of modern Internet usage. It is this context that compels the conclusion that an average Internet user would not be confused and would know that a .com website is probably a commercial website while a .cam websites would be something more focused in a particular field.

In AC Webconnecting, he wrote:

I agree that a consumer would quickly realize that a .cam website is likely associated with photography or camera use and is different than a .com website in use generally by a myriad of commercial entities.

So he’s putting the “greater emphasis” on usage — a factor that is not explicitly mentioned in the Applicant Guidebook’s description of the SCO and which may quite often differ between applicants.
Right there, in Smith’s interpretation of his task, we have a reason why SCOs will produce different results for identical strings.
I find Smith’s thinking baffling for a couple of reasons.
First, “a consumer would quickly realize that a .cam website is likely associated with photography” seems to ignore the existence of a bazillion .com web sites that are also associated with photography.
When did “commercial entities” and “photography or camera use” become mutually exclusive? Is photographyblog.com not confusingly similar to photographyblog.cam?
Second, he ignores the fact that basically anyone will be able to register a .cam web site for basically any purpose. None of the applicants want to restrict the gTLD to camera-related stuff.
ICDR panelist Laeuchli, in the Demand Media .cam case, raised this precise point, saying:

“.com” and “.cam” would use the same channels appealing to a broad audience. Even though according to Applicant, its envisioned TLD will “likely appeal” to a specific audience, it plans to operate “.cam” as an open gTLD. This would lead to extensive overlap.

Panelist Smith has some other notions about confusion that seem to defy common sense. He wrote in the AC Webconnecting case:

The .com TLD is the most widely recognized string in the Internet world. No reasonable Internet user would fail to recognize the .com TLD. The very reputation of the .com name serves to limit the potential for an average Internet user to be confused by the proposed .cam TLD. It is indeed unlikely that an online consumer would confuse a .com website with a .cam website.

Does this not strike anyone else as bad thinking?
It seems to me to be a little like saying that it’s perfectly okay to market a brand of carbonated beverage called Cuke, because Coke is so famous that nobody could possibly be confused. I don’t know where the law stands on that issue, but I’m pretty sure Coke wouldn’t be happy about it.
There’s also some weirdness in Laeuchli’s decision in the Demand case.
He puts some weight on the similarity scores produced by the controversial Sword algorithm in his decision, but apparently without doing even the basic research. He writes in his findings:

No matter what the standards and purpose the ICANN SWORD algorithm includes, it has comparative value.
…
Since pairs such as “God” and “dog” (85%) reach similarity scores of 84% and higher, how much more similar would “cxm” and “cxm” be (x being replaced with a vowel)!

The answer is that, according to Sword, they’re less similar. Sword scores “cam” v “com” at 63%.
Laeuchli knows it’s 63%, because he makes reference to that fact in his summary of Verisign’s evidence. He doesn’t need to speculate about the number based on what “god” v “dog” scores (and if he did the “dog” v “god” query himself, why on earth didn’t he just query “com” v “cam” too?)
His finding that .cam and .com will cause probable confusion seems to be based largely on expert witness testimony provided by both Verisign and Demand, in which he found Verisign’s more persuasive.
This evidence seems to have largely comprised the opinions of linguists, examining mouth shapes and acoustic frequencies, and market research looking into internet user behavior. As none of it has been published, it’s difficult to judge which side had the better arguments.
But it’s undeniably about the similarity of the strings, rather than the proposed usage, which makes Demand Media’s statement today — that SCOs “are meant to be applicant agnostic and have nothing to do with the registration or use of the new gTLD” — quite confusing.
Demand lost its case based on the string similarity, whereas the other two applicants won theirs based on the usage.
Perhaps Demand senses that its .cam application will not be immediately rejected if ICANN reopens the debate about string similarity. If think it’s probably correct.

Google beats Donuts in objection — .pet and .pets ARE confusingly similar

Google has won a String Confusion Objection against rival new gTLD applicant Donuts, potentially forcing .pet and .pets into the same contention set.
The shock ruling by International Centre for Dispute Resolution panelist Richard Page goes against previous decisions finding singulars and plurals not confusingly similar.
In the 11-page decision, Page said he decided to not consider the reams of UDRP precedent or US trademark law submitted by the two companies, and seems to have come to his opinion based on a few simple facts:

Objector has come forward with the following evidence for visual, aural and meaning similarity. Visually, the words are identical but for the mere addition of the letter “s”. Aurally, the word “pets” is essentially phonetically equivalent to the word “pet”. The term “pet” is pronounced as it is spelled, “pet”. The term “pets” is likewise pronounced as “pets” in essentially a phonetically equivalent fashion. The terms each have only one syllable, and they have the same stress pattern, with primary accent on the initial “pe” portion of the words. In commercial meaning, the terms show no material difference. As English nouns, “pets” is the pluralization of “pet”.
The visual similarity and algorithmic score are high, the aural similarity is high, the meaning similarity is high. Objector has met its burden of proof. The cumulative impact of these factors is such that the Expert determines that the delegation of <.pet> gTLD and the <.pets> gTLD into the root zone will cause a probability of confusion.

Page did take into account the similarity score provided by the Sword algorithm — for .pet and .pets it’s actually a fairly weak 72% — in his thinking on visual similarity.
But he specifically rejected Donuts’ defense that co-existence of plurals at the second level was proof that plural/singular gTLDs could also co-exist at the top-level, saying:

The rapid historical development of the Internet and the proliferation of domain names over the past two decades has taken place without the application of the string confusion standard now established for gTLDs. Therefore, the Expert has not considered the current coexistence of pluralized second-level TLDs or similarities between country code TLDs and existing gTLDs in the application of the string confusion standard in this proceeding.

Can: open. Worms: everywhere.
The decision stands in stark contrast to the decision (pdf) of Bruce Belding in the .hotel v .hotels case, in which it was found that the two strings were “sufficiently visually and audibly different”.
Likewise, the panelist in .car v .cars (pdf) found that Google had not met the high evidential bar to proving the “probability” rather than mere “possibility” of confusion.
One has to assume that the evidence Google submitted in .car is fairly similar to the evidence it submitted in .pets.
Are String Confusion Objections just a crap shoot, the outcome depending on which panelist you get? It’s probably too early to say for sure, but it’s looking like a possibility.
The big test will come with the next .pets decision. Afilias, the other .pet applicant, has also filed an SCO against Donuts over its .pets bid.
What if the panel in the Afilias case goes the other way? Will Donuts be in a contention set with Google and Afilias or won’t it?
I asked Akram Atallah, president of ICANN’s Generic Domains Division, about this yesterday and he said that ICANN basically doesn’t know, and that it might have to refer back to the community for advice.
Read the Atallah interview here and the .pets decision (pdf) here.

First string confusion decisions handed down, Verisign loses against .tvs

The International Centre for Dispute Resolution has started delivering its decisions in new gTLD String Confusion Objections, and we can report that Verisign has lost at least one case.
ICDR expert Stephen Strick delivered a brief, five-page ruling in the case of Verisign vs. T V Sundram Iyengar & Sons yesterday, ruling that .tvs is not confusingly similar to .tv.
TVS is a $6-billion-a-year, 100-year-old Indian conglomerate, while .tv is the ccTLD for Tuvalu, which Verisign manages because of its similarity of meaning to “television”.
It’s impossible to glean from the decision (pdf) what Verisign’s argument comprised. The summary is just two sentences long.
But TVS, in response, appears to have relied to an extent on the “DuPont factors” a 13-point test for trademark confusion that came out of a 1973 case in the US.
That’s the same precedent that has been found relevant in many Legal Rights Objections in cases handled by WIPO.
The “discussion and reasons for determination” section of the .tvs decision, in which Strick found that confusion was possible but not “probable”, amounts to just four sentences.
Here’s almost all of it. Emphasis in original:

in order for the Objector to prevail, Objector must prove that the co-existence of the two TLDs in question would probably result in user confusion. Given the analysis of the thirteen factors cited by Applicant derived from the DuPont case cited above, I find that Objector has failed to meet its burden of proof regarding the probability of such confusion. I note that while the co-existence of the two TLDs that are the subject of this proceeding may result in confusion by users, Objector has failed to meet its burden of proof to establish the likelihood or probability that users will be confused.
In considering parties’ arguments, I was persuaded, in part, by Applicant’s arguments relating to the commercial impression of the TVS TLD, including the proof offered by Applicant as to the longevity of the TVS brand, the limited nature of the gTLD’s intended use, the dissimilarity of the goods or services associated respectively with the two strings, ie TVS’s association with automobile products, the fact that TVS’s brand is associated with capital letters (whereas Objector’s .tv is in lower case), the fact that TVS is well known and associated with its companys’ [sic] brands, the lengthy market interface and the long historical co-existence of TVs and tv without evidence of confusion in the marketplace.

The geeks among you will no doubt be screaming at your screen right now: “WTF? He thought CASE was relevant?”
Yes, apparently the fact that the TVS trademark is in upper case makes a difference, despite the fact that the DNS is completely case-insensitive. Bit of a head-scratcher.
I understand several more decisions have also been sent to applicants and objectors, but they’re not yet pubicly available.
The ICDR’s web site for new gTLD decisions has been down for several days, returning 404 errors.