Re: [regext] rfc7484bis

[Marc Blanchet <[e-mail]>]

On 11 Aug 2020, at 15:27, Patrick Mevzek wrote:

> Hello Marc,
>
> On Tue, Aug 11, 2020, at 13:55, Marc Blanchet wrote:
>> On 4 Aug 2020, at 15:47, Patrick Mevzek wrote:
>>>
>>> PS: related but not directly, at least for domain registries, it 
>>> would
>>> be
>>> nice to have an `SRV` record on `_rdap._tcp` or something to point 
>>> to
>>> relevant
>>> RDAP server, even if that does not allow to encode the path (but 
>>> maybe
>>> a solution with .well-known/ and URI template could be found), it
>>> allows at least
>>> for nice failover and load balancing. It may be a problem for gTLDs 
>>> as
>>> they have
>>> restrictions in content of their zone.
>>
>> well, this has been debated at length during the WEIRDS working group
>> work. I actually wrote a sentence about this in the RFC (in the
>> acknowledgements section). I’m not sure we want to restart the 
>> debate
>> again…
>
> I am not saying to restart the debate, especially not in the context 
> of a -bis document where protocol changes are not welcome.
>
> But the RFCs are also 5 years old now and a lot of things change 
> quickly.
> SVCB record in the DNS being one, while not there already.
>
>>> Maybe the newly expected SCVB record could help...
>>>
>>> A setup like that would allow for discoverability without
>>> centralization of data,
>>> which also removes IANA from the hot operational path when RDAP
>>> clients do queries.
>>
>> yes. this is the well-known caveat of this RFC and discussed and 
>> debated
>> during WEIRDS. But experience up to now has not shown any issue, at
>> least to my knowledge. (and as a developer of the RDAP Browser mobile
>> app, I haven’t seen any issue fetching that registry. I do have 
>> found
>> thousands of issues with the registry/registrars RDAP servers 
>> however,
>> but that is another story).
>
> I am not saying there is a current issue, fetching the JSON file from 
> IANA
> webserver is clearly the smallest problem of any RDAP client.
>
> But I also think there is currently no issue because basically the 
> world did
> not shift to RDAP in any way yet. Which can easily be witnessed by the 
> amount
> of broken servers so far - even if they are in a regulated space where 
> compliance
> is an issue - and the total lack of ccTLDs in this space, at least in 
> operations.
>
> Once there is a shift, then an issue can happen. Or not. We can indeed 
> not
> try to prematurely optimize this.
>
> And I would prefer a more decentralized way of discovery, because it 
> makes
> more sense to me and because we do already have a decentralized 
> publicly available
> database (the DNS) that could store RDAP related information per 
> registry.

we all wanted. But when one think of all the requirements and that we 
make sure it works for both domains and IP addresses and AS numbers, the 
best compromise solution was the IANA registries. None of the solutions 
discussed were perfect. Yes, the IANA registry solution is not 
distributed, but brings additional features (as agile as we want with a 
JSON data model) that were not available in in-dns solutions, available, 
ready to deploy, and also, one of the point that was pushed often was 
the ability to fetch the data from a contained Javascript client in a 
browser (where dns queries are not available).

Marc.

>
> -- 
>   Patrick Mevzek
>   [e-mail]