Re: [regext] New Version Notification for draft-loffredo-regext-epp-over-http-02.txt

I agree that a redesign would only make sense if it can do more afterward than before. Otherwise, it would only produce costs and effort without delivering any real added value.

I may not doubt whether the attractiveness can be an added value, but I would at least question whether that is a sufficient added value.

In my years as a registrar, missing or unequal test systems and detailed documentation (e.g., error codes, behavior, poll messages, limits, etc.) were a much bigger problem than how the transport works—especially when transferring TLDs from one backend to another.

I don't know how your experiences are/were in those cases, but if I had a choice, I would instead work on simplification/documentation. If I remember correctly, there was also once an approach to this from Jim https://datatracker.ietf.org/doc/draft-gould-carney-regext-registry/ <https://datatracker.ietf.org/doc/draft-gould-carney-regext-registry/>

Regarding Mario's draft, EPP over HTTP is already in use, and the document, at least as I understand it, is only intended to standardize the existing procedure, so there is no proliferation. I support that.

Tobias

> On 21. Jun 2022, at 22:51, Patrick Mevzek <pm@dotandco.com> wrote:
> 
> On Tue, Jun 21, 2022, at 15:09, Eduardo Duarte wrote:
>> EPP is about 20 years old and I 
>> think it needs some reshaping to the actual Internet state. 
> 
> Is the EPP *transport* really where people struggle most? Ok, XML over TLS might not be the current trendy couple in Internet circles, but among all the problems I see in EPP (having worked both on registrar and registry side), I really do not have the transport in my top 10.
> The plethora of extensions to do the same thing, and the various "quality" of extensions is more of a concern to me, as well as the non-existent discoverability of features (there is one extension solving part of the problem, not used by everyone). Or the lack of standardization in error codes/messages/details extended from core case. Or the now not good enough design of a contact. Or operations being mixed where they shouldn't (like restore in update).
> 
> We barely arrived only a few months ago to have "fees" finally being a standard... and it
> will take years before all registries switch to it. This is certainly where registrars struggle more than just having to use a TLS "socket" (I count around 28 versions of the fees document, for 18 different XML namespaces with more than a couple of them really used in the wild).
> 
> Said differently, the transport part seems to me really the easy part of the problem of EPP viewed globally. Of course, if that blocks some actors, then the working group is certainly the relevant place to find out a standardized solution, but will really a lot of registries suddenly switch to it just for the sake of switching to it?
> 
> Or: what EPP over TCP and/or a REST redesign really add as new features, solving current problems? Besides "it looks similar to the rest of the Internet, so it will attract more/better programmers" (a statement I would certainly have an hard time to believe) or "the crappy hosting I am using only allows HTTPS servers/clients and nothing else, so now I have to adapt my whole word just because I chose a bad system from the beginning".
> 
> Just my personal views of course.
> 
> -- 
>  Patrick Mevzek
>  pm@dotandco.com
> 
> _______________________________________________
> regext mailing list
> regext@ietf.org
> https://www.ietf.org/mailman/listinfo/regext