2022-04-28 IDNs EPDP - Meeting #33

Notes/ Action Items

Action Items:

Action Item 1: Staff to request volunteers to join a small group tasked with developing examples related to Levels 1, 2, and 3. Staff to circulate a doodle to find a time for the small group to meet.

Notes – IDNs EPDP Call – 28 April 2022

Welcome & Chair Updates (5 minutes)

• On today’s call EPDP Team members will be able to explain their responses on the mailing list regarding the three “levels” so that the group can try to converge on an approach to Charter Questions E3, E1 (Part 1), and E3a.

Charter Questions E3, E1 (part 1), E3a – Continued Discussion of the Three Levels (50 min)

• Slide 4 – overview of Charter Questions E3, E1 (part 1), E3a
• Slide 5 – Comparison matrix – explanatory notes
• Slide 6 – Comparison matrix – consolidated view
• Slide 7 – Comparison matrix – factors for consideration
• Comment: In Level 1, all entities will know that they can get the strings they apply for and if they apply in the future, they may not be able to get them based on the review for confusing similarity. The EPDP Team member disagrees with the consequences listed on the slide. Level 1, 2, and 3 all guarantee that there will be no confusability. None provides better protection than another in this regard.
• Slide 8 – Level 1: Primary + ONLY requested allocatable variants
• Comment: Support this option if variants can only be requested inside application rounds. This is not a contribution on behalf of RySG. RySG does not have a position on this topic. This level is sufficient because it is a safeguard against allocating variants that would be confusable to other strings that are allocated. There is no need to check any future potential confusability of labels that may never be applied for or allocated. That would only restrict the space without any really necessity.
• Comment: Support for the above comment. No need to restrict the space without reason. Level 1 does not restrict the space without reason, entails the least similarity work, and involves the lowest cost. The only drawback is that if an entity does not apply for an allocatable variant initially it might not be able to get it later. But if the applicant knows this from the beginning, it can apply for the variants it wants at the outset. All three levels are the same in terms on avoiding confusability. Whether level 1 or 2 is better depends on costs associated and whether applicants can activate strings only within rounds or at any time.
• Slide 9 – Primary + ALL Allocatable Variants
• Comment: Support this option if variants can be requested outside of application rounds. If a registry can request any allocatable variant at any time, level 1 is not enough. In this case, at the application round of the originally applied for TLD we need to check all of the allocatable variants. We need to ensure from the start that these allocations will be possible.
• Comment: Support for level 2 because it narrows the chances of confusability in the same round which improves predictability. It presents a good balance from a cost perspective.
• Comment: Support because it provides flexibility and opportunity while ensuring a degree of protection of confusability.
• Comment: Some support depending on other elements such as costs associated with each string applied for and when a string could be applied for. Allows entity to ensure that all of their allocatable strings and all other possible future similar strings would be blocked. This creates predictability because all entities have their original string and allocatable variants. No other TLD in the future could apply for those allocatable variants. This is a good middle ground.
• Comment: ALAC supports Level 1 if costs associated with the applied for string is high. In such a case Level 2 doesn’t make sense. If the cost is relatively low, Level 2 is more appropriate. From the perspective of confusability, they all provide the same level of protection.
• Slide 10 – Level 3: Primary + ALL Variants
• Comment: Support for level 3, with continuous technological improvements in improvements in reducing Risk Evaluation time of assessment. Suggestion to use AI to support the evaluation. The EPDP Team previously discussed that AI may not be able to be leveraged for this purpose.
• Comment: There are many complaints about misuse of domain (mainly .brands). String similarity can cause more issues.
• Comment: We might also need to think about if we go with Level 1 or 2, what does first come first serve mean when prior round applicants activate variants. In the case of Level 3, that would not be an issue. In the case of 1 and 2 that would be an extra point of consideration.
• Comment: In Level 2 all would have been taken care of as well.
• Comment: We have been looking at this from the perspective of applicants, but we also want to make the space less confusing from an end user perspective. This is the underlying motivation of string similarity review. Examples: At the second level, if we have myname.com vs. myname.org there is no confusability. If we replace .com and .org with strings that are similar – myname.tld vs. myname.tld’ where tld’ is very similar to .tld, this might be confusing for the end user. If myname.tld is similar to a variant of another TLD that is delegated there could still be confusability from an end user perspective through a transitive relationship. Irrespective to whether the variant is allocatable, blocked, etc, there can still be confusability when there are two similar strings in the DNS space.
• Example: Simplified and traditional Chinese – in most cases the two are visually distinct. If applicant A applies for a simplified Chinese version, which has a traditional version that looks different but is considered to be the same in meaning and has not been applied for. If applicant B applies for a traditional Chinese character that is visually very similar to traditional variants of the simplified character applied for by applicant A. If the comparison is done only for applied-for strings, the applied-for strings for applicant A and b will both be delegated, the end users can get confused by seeing the traditional version is the same as the simplified version. This will create confusion for all registrations under those TLDs. Level 1 will not prevent this scenario.
• Comment: Preference for a combination of level 1 and level 3. For string similarity review there are two things to compare, Set A and Set B. If it includes variants there is Set A and Set B. Set A is the strings we would like to compare. Set B is the strings we compare against. Set A includes primary + only requested allocatable variants. Set B it includes reserved names, existing TLDs + All variants (allocatable and blocked variants), strings requested as IDN ccTLDs and all variants allocatable and blocked, other applied for gTLDs and all variants allocatable and blocked. The core of the problem is whether we allow applicants to request variants outside of a round. We should focus on Set B, which can decide what is similar or not. To meet the conservativism principle, Level 3 is best. Therefore, Set 1 is Level 1 and Set B is Level 3.
• Discussion
• Comment: Confusing similarity has always been only about visual similarity. Applying that principle, the maximum conservative approach is not consistent with the way we have always approached these things. How would this look in the case of variants that are not visually similar? How would this be any different from a case of two strings that have the same meaning but look different? It’s hard to compare these levels without examples. Why would we give additional protections here?
• Response: The difference is 1. color vs. colour and 2. color vs COLOR. The example above is talking about case 1, IDN Variants is closer to case 2, whereas in English alphanumeric is how we deal with it today it is 'the same' but with IDN, we have a situation of 2. that is technically not the same but linquistically the same.
• Comment: We have already acknowledged that the treatment of IDNs and variants is different from other strings.
• Comment: That is the case within a variant set, but it is another level when we are talking about similarity across variant sets.
• Comment: The primary position from ALAC is Level 2, but this position is conditional and some additional discussion might be needed. Examples would be helpful to support additional deliberations on this topic.
• Question: If we reached an agreement on whether when you apply for the IDN and apply for the variants you want, it is done in a round, does this overcome any of the differences in this discussion?
• Response: No, this would not necessarily settle the issue.
• Comment: Agree that this would not resolve the issue. But if you go with Level 2 it should be possible to activate variants between rounds.
• Comment: We should think about activation of variants like updating nameserver or updating the backend. It should involve technical checks but should be possible at any time. Allocating in rounds is not a good starting point. We should look at actual examples that address these issues visually. Maybe a small group should put some examples together.
• Comment: Agree that this should be the case if you are activating a variant you already expressed interest for in an application. Disagree if the registry has not previously expressed interest in through an application. This is consistent with the SSAC advice, because applicants will have to explain how they can handle the variants and these statements need to evaluated.

Action Item 1: Staff to request volunteers to join a small group tasked with developing examples related to Levels 1, 2, and 3. Staff to circulate a doodle to find a time for the small group to meet.

AOB

• Reminder to provide feedback on the outreach letter to GPs by Friday. It has been shared on the mailing list.