Thanks Patrick for the full history on templates and groups
In mean time I looked to the .cz extension and it seems they serve the same
purpose.
I think the 2 flavours for nameservers should not be an obstacle. When using a
template, it would be possible to use either host links vs host attributes. It
depends on how you design it.
I see a big difference with bulk updates. Bulk updates equals to lots of data
to be transferred.
With templates, you further "normalize" the data model. It is comparable with
contact updates: when you change the name of a contact, all WHOIS data of
domains having the same contact id are updated with a single update command.
When you change a template by updating nameservers/DNSSEC data/..., all DNS
data of domains having the same template id are updated with a single command.
The client is relieved of the burden to check whether all domains with the same
DNS "profile" are updated and it does not have to perform a bulk update of
thousands of domains.
I'm also not in favour of a protocol extension for bulk updates. The reason is
that it is already possible. A client may handle updates asynchronously and can
shoot a list of command to the EPP server.
Kind regards
Pieter
> On 26 May 2018, at 04:48, Patrick Mevzek wrote:
>
> Pieter,
>
> On Fri, May 25, 2018, at 21:37, Pieter Vandepitte wrote:
>> The registry I work for, developed a custom extension to manage
>> "nameserver" groups and "keygroups". When a registrar links a group to a
>> domain, all member nameservers/keys of that group are automatically
>> linked to that domain. This way, it is very convenient for DNS operators
>> to update DNS data on their complete domain portfolio with a single
>> group update, without forgetting a domain.
>>
>> It is used quite a lot, but I did not find other registries having this
>> kind of functionality (I did not perform an extensive search). I'm quite
>> sure we are not the only ones, so do you know other registries having
>> this?
>
> I know two extensions:
>
> - the .BE/.EU one: the last time I have looked at them, it was the same,
> except for the namespace
> - the .CZ one, in fact in their Fred open source EPP server (so probably used
> for their others TLDs), see
> https://fred.nic.cz/documentation/html/EPPReference/ManagedObjects/Nssets.html
> and
> https://fred.nic.cz/documentation/html/EPPReference/ManagedObjects/Keysets.html
>
>
>> From memory, they cater for the same needs, but are basically incompatible,
>> besides the namespace.
> Starting with the terminology: "group" on one side, "set" on the other.
>
> I would add 2 remarks:
> - for nameservers it seems to me to make more sense when hosts are objects,
> instead of attributes while obviously it works in both case. The world is
> quite divided on this, gTLDs are mostly (only?) in the objects group, while
> ccTLDs are predominantly in the attributes group
> - for DNSSEC material, here we hit another problem, the DS vs DNSKEY
> dichotomy, partially reflected in the dsData vs keyData interfaces of
> secDNS-1.1
> All grouping cases only make sense of course with the DNSKEY case, because
> the DS depends on the domain. Again, without any hard facts, I still believe
> that most registries are using the DS case. Some, even with the dsData
> interface, ask also for the underlying keyData, but only to check that the DS
> was correctly computed.
>
> Also, such kind of features have consequences for transfers.
>
>> Is there any interest from the community to offer such a feature to
>> their registrars and collaborate on a common extension? I think of
>> something more generic in a way that a registrar can create a "template"
>> for any kind of object and apply that template to other objects. This
>> way, besides the benefits for DNS operators, a registrar could also
>> define e.g. a default admin contact for every domain, or even apply
>> custom extensions to every domain…
>
> Note that there were various attempts to define features such as templates,
> containers, or bulk operations.
>
> Specifically for bulk operations, since the discussions often circled around
> that the primary argument was that EPP is a provisioning protocol and as such
> is not tailored for bulk operations. Which brings immediately this counter
> argument: ... but you can query for more than one object in a given
> command.
>
> Note that while not completely the same, issues of "bundling" domain names
> due to IDN variants typically is also loosely related to all of this.
>
> One of the latest iteration around these concepts was this draft:
> https://tools.ietf.org/html/draft-gould-regext-dataset-01
>
> HTH,
>
> --
> Patrick Mevzek
>
> ___
> regext mailing list
> regext@ietf.org
> https://www.ietf.org/mailman/listinfo/regext
___
regext mailing list
regext@ietf.org