Patrick, I respond to your comments below.
-- JG James Gould Distinguished Engineer jgo...@verisign.com <applewebdata://13890C55-AAE8-4BF3-A6CE-B4BA42740803/jgo...@verisign.com> 703-948-3271 12061 Bluemont Way Reston, VA 20190 Verisign.com <http://verisigninc.com/> On 1/28/20, 2:29 AM, "regext on behalf of Patrick Mevzek" <regext-boun...@ietf.org on behalf of p...@dotandco.com> wrote: On Fri, Jan 24, 2020, at 09:51, Antoin Verschuren wrote: > This is a formal adoption request for Extensible Provisioning Protocol > (EPP) Secure Authorization Information for Transfer: > https://datatracker.ietf.org/doc/draft-gould-regext-secure-authinfo-transfer/ > > Please review this draft to see if you think it is suitable for > adoption by REGEXT, and comment to the list, clearly stating your view. While the subject of transfers between registrars might need work, I am not convinced this working group is the appropriate place for that, as in particular this document seems to mostly impose directives on registries and how they should handle and store domain passwords which is for me off topic for EPP as a protocol, and transfer issues cover points that are not completely technical but also business related, including the specific security model in which we want to work. (one might note that multiple procedure specify a transfer "undo" procedure, while this does not exist technically anywhere in EPP land...) JG - You have brought this up before, but the adoption of a BCP draft by the WG is certainly applicable. A transfer "undo" procedure is something new and can be handled separately. As discussed in other threads, I am more leaning towards a ground up discussion on passwords attached to domains, and if other models can exist here. So I think work should instead go in that direction, which is then really completely EPP related, while discussions on passwords size, complexity, entropy, storage, TTLs, and so on as found in this draft are for me clearly out of scope of both the working group and EPP related specifications. The authInfo node was defined from the ground up to be extensible, and we should leverage that to put into place better mechanisms than current plain text passwords. JG - An alternative approach is not mutually exclusive and I look forward to any proposals that you may have. I also fear discussions may forget there are already today other models than the "common" gTLD one that the draft tries to change, and I would like to make sure they are taken into account when drafting a solution. Among others, some registries use a "push" transfer model (so no passwords needed at all any more) and other registries basically do not let registrars set/handle passwords anymore, the first step of a transfer is the loosing registrar to ask the registry to generate a password that is in fact directly sent to the registrant, who in turn will input it at the gaining registrar (with some time window limits). JG - Any input from interested parties is welcome. The main question is whether other models can be fit into a common practice or whether they stand on their own. Also while things should be left separate to be really able to produce new ideas, transfer issues can not be tackled without at least looking a little around RDDS and specifically RDAP, and around laws such as the GDPR from EU land, because this has the direct consequence for example that some registries do not continue to collect contacts, or none besides the registrant. JG - One goal of enhancing the security of the authorization information for transfers is to reduce or remove the dependency on the use of RDDS. PS: as an exercise I reviewed how a batch of registries currently reply to domain:info queries in the following case: - registrar is not sponsoring registrar, and not providing authInfo - registrar is not sponsoring registrar, and providing invalid authInfo - registrar is not sponsoring registrar, and providing correct authInfo - domain does not exist and registrar is providing authInfo - domain does not exist and registrar is not providing authInfo Results vary a lot, even when just looking at the EPP return code. Also the content of <infData> can change, and differ - even outside of the password - between sponsoring and non sponsoring registrars. There is clearly no standardization here and this directly impacts how transfers can be done by registrars (on issues for example of being able to test the password without really starting the transfer, or to know the current nameservers attached to the domain, or its expiration, etc.) JG - draft-gould-regext-secure-authinfo-transfer focuses on the behavior of the info command and response as it relates to verifying the authorization information value and disclosing whether the authorization information is set in the info response to the sponsoring or non-sponsoring registrar. Defining a BCP would help to make things more consistent. draft-gould-regext-secure-authinfo-transfer does not cover what elements are returned in a full or partial info response. My thought is that if the valid authorization information is provided by a non-sponsoring registrar that the full info response should be returned. Something like that could be added to draft-gould-regext-secure-authinfo-transfer if it would help with the transfer. Your thoughts on this would be helpful. -- Patrick Mevzek p...@dotandco.com _______________________________________________ regext mailing list regext@ietf.org https://www.ietf.org/mailman/listinfo/regext _______________________________________________ regext mailing list regext@ietf.org https://www.ietf.org/mailman/listinfo/regext
