Re: [scim] Multiple external id values

[Mark Dobrinic <[e-mail]>]

In our particular case, you can indeed see that the client acts as a
hub, applying multiple external identities for multiple domains.

I totally agree that is is generally a bad idea to expose an identity
that is registered with a client for domain A, to a client for domain B.
Not for minimizing PII exposure primarily, but for exposing information
that weakens the linkability property of privacy with identifiers.

This is not what we try to solve though. It is that first case. We'd
like to use Scim towards a backend system for the client-as-hub-case.
The externalId attribute can not help with that.

That's why I try to raise a general solution that does.

Putting it in an extension as described is totally acceptable though. I
don't feel like disturbing the choices that were made through consensus,
but I do feel like coming up with a general solution for that
alternative usecase. Including possible privacy implications.


On 05/10/15 17:55, Phil Hunt wrote:
> I did raise this issue. 
>
> The consensus was one and one only. 
>
> I do have a concern about multiple domains connecting to the same entity. 
>
> Does one see this as a client hub controlling multiple cloud spokes.
> Or does one see this as multiple client spokes contributing to a
> common hub entity. 
>
> The latter has PII propagation concerns. 
>
> I suspect most for now only care about the former. 
>
> Phil
>
> On Oct 5, 2015, at 08:39, Mark Dobrinic
> <[e-mail]
> <mailto:[e-mail]>> wrote:
>
>> Hi Phil,
>>
>> Thanks for the quick response.
>>
>> The resolution to expose a client specific externalId only works in
>> half the scenario's, where you have multiple clients all representing
>> one domain. In the other scenario, where one client representing
>> multiple domains is using one Scim Service Provider, the client
>> actually relies on the Service Provider to store multiple identities
>> for its multiple domains, so in that usecase it is multi-valued by
>> design.
>>
>> Do I get it right that this particular use case has not yet come up then?
>>
>> Concerning the privacy cinsideration, given a trust relationship
>> between the Scim Service Provider and the Scim Client (quite some
>> features are based on this already), privacy characteristics of the
>> solution are not of primary concern here. The Scim Client could
>> (should) actually take over the responsibility for protecting privacy
>> in this one.
>>
>> Cheers!
>>
>> Mark
>>
>>
>>
>>
>> On 05/10/15 17:17, Phil Hunt wrote:
>>> This was discussed and the consensus was single value. 
>>>
>>> What you can do is expose a different single value for each domain
>>> client your server talks to.  Thus domains cannot see each other's
>>> ids. The spec already suggests this as a security consideration for
>>> the id value. 
>>>
>>> This would provide much better privacy characteristics and reduces
>>> propagation of PII. 
>>>
>>> Phil
>>>
>>> On Oct 5, 2015, at 07:42, Mark Dobrinic
>>> <[e-mail]> wrote:
>>>
>>>> Hi Scim people,
>>>>
>>>> While incorporating Scim client functionality towards a Scim
>>>> Service Provider, we've stumbled upon a limitation regarding our
>>>> desired use of externalId, that doesn't fit the current spec.
>>>>
>>>> Our usecase involves managing multiple external id's, so that we
>>>> can both support multiple externalId's to multiple Scim Clients
>>>> from one Scim Service Provider, as well as let a Scim Client manage
>>>> multiple external id's with one Scim Service Provider.  Each
>>>> external id lives within the domain of a, well, domain. There are
>>>> plenty of real world examples where domain-scoped id's are used, so
>>>> it seems no more than natural to consider this with Scim as well.
>>>>
>>>> For example, a user with id barbara has an externalId 'babs' in the
>>>> domain 'home', and externalId 'labarbara' in domain 'work'.
>>>>
>>>> As I understand Scim as being as effective and simple as possible,
>>>> we've been using the extensibility of Scim to make this work. I can
>>>> imagine this to be a generic problem that more people are (or will
>>>> be) facing. Therefore, I'd like to propose our solution and would
>>>> like to discuss whether this could/should be considered a standard
>>>> solution to a standard problem, either in this form or something
>>>> better.
>>>>
>>>> What we did was introduce an extension "
>>>> urn:scim:schemas:extensions:external-ids:1.0"
>>>> that defines an attribute "externalIds",
>>>> is of type "complex",
>>>> is multiValued,
>>>> can be described as "External Identities of a user. Each External
>>>> Identity is scoped within its domain",
>>>> is optional,
>>>> and has the following sub-attributes:
>>>> - value : string value containing the value of the external id
>>>> - type : string value containing the domain of the external id
>>>>
>>>> In JSON, it would look like this:
>>>>
>>>>                 {
>>>>                 "schemas": [ "urn:scim:schemas:core:1.0",
>>>> "urn:scim:schemas:extensions:external-ids:1.0" ],
>>>>                 "id":"2819c223-7f76-453a-919d-413861904646",
>>>>                 "userName":"bjensen",
>>>>                 "externalId":"bjensen",
>>>>                 ....
>>>>                 "urn:scim:schemas:extensions:external-ids:1.0": {
>>>>                     "externalIds": [
>>>>                         {
>>>>                             "value": "babs",
>>>>                             "type": "home"
>>>>                         },
>>>>                         {
>>>>                             "value": "labarbara",
>>>>                             "type": "work"
>>>>                         }]
>>>>                     }
>>>>                 }
>>>>
>>>>
>>>>
>>>> Note that we are adding to the baseline features of Scim, and not
>>>> changing anything existing.
>>>>
>>>>
>>>> I would like to hear your feedback on this, and I'm interested
>>>> whether this could be considered as a standard extension in
>>>> upcoming Scim versions.
>>>>
>>>>
>>>> Thanks!
>>>>
>>>> -- 
>>>> Regards,
>>>>
>>>> Mark Dobrinic
>>>> Software Engineer and Identity Specialist
>>>> Twobo Technologies AB
>>>>
>>>> [e-mail]
>>>> www.twobotechnologies.com
>>>> _______________________________________________
>>>> scim mailing list
>>>> [e-mail] <mailto:[e-mail]>
>>>> https://www.ietf.org/mailman/listinfo/scim
>>
>>
>> -- 
>> Regards,
>>
>> Mark Dobrinic
>> Software Engineer and Identity Specialist
>> Twobo Technologies AB
>>
>> [e-mail]
>> www.twobotechnologies.com
>> _______________________________________________
>> scim mailing list
>> [e-mail] <mailto:[e-mail]>
>> https://www.ietf.org/mailman/listinfo/scim


-- 
Regards,

Mark Dobrinic
Software Engineer and Identity Specialist
Twobo Technologies AB

[e-mail]
www.twobotechnologies.com