Re: [scim] [EXTERNAL] Query on a specific known resource

[Phillip Hunt <[e-mail]>]

Julien,

I believe the answer was that whenever a “filter” is supplied, the response MUST (as in normative) be in the form of a ListResponse unless an error has occurred. Returning the resource is not valid.

I don’t believe your suggestion would qualify as an Errata (but you can still submit it). This issue has a “wouldn’t it have been simpler if SCIM just did this” feel.  Such a change would be “normative” change to the protocol rather than an Errata.

Chairs, please correct me appropriately if I am wrong…

IETF RFCs can’t be directly revised this way even if the consensus of the group has changed since RFC7644 was published. The correct process would be the working group discussing changing behavior in a future SCIM version (meaning a new RFC and protocol version). 

Is just returning a resource for a query against the resource really simple?
In my opinion no. The simpifying assumption only handles the “match” case and fails to consider negative cases. If such a change would made, SCIM Service Providers would lose the ability to signal filter NOT MATCHED vs. NOT FOUND (404) as servers would only be able to respond with Status 404.   Yes, instead of 404 we could still return a ListResponse, but now you are making single resource queries signalling very complex.  As written now, a query returns a ListResponse unless there is an error (no match is not an error).

—> again…this is all debatable as to what way is right. It is something for the group to discuss in a future draft.  I am not that against this. It’s just that there is a process. :-)

As practical advice, I would assume mistakes in implementation are made.  Make sure your client parser looks at the “schemas” attribute to see what you have. If you don’t find the “ListResponse” schema, you can then check to see if it is a resource.  That’s one of the “failsafe" features of SCIM that allows parsers to figure out what was received by looking at the “schemas” attribute.

Phil Hunt
Editor for RFC7644

> On Jul 15, 2022, at 6:14 AM, Julien Schneider <[e-mail]> wrote:
> 
>  Hi all,
> 
> I didn't get any answers to my previous mail, so I guess my thinking and interpretation are more or less correct. 
> 
> I still think that the RFC is a bit confusing on that topic. I'd probably like to shortly discuss the issue with the working group to decide if I should file an errata for this. Unfortunately, I wont be able to participate to IETF 114, but my colleagues Joris and Hans-Jorg (in CC) will participate in person. 
> Joris is aware of all that and should have some time to talk about it in the side meetings if that makes sense for some of you (I guess side meetings might be sufficient/more appropriate for this kind of things?).
> 
> Thanks again,
> Julien Schneider
> Tel: [phone]
> Fax: [phone]
> 
> http://www.audriga.com <http://www.audriga.com/> | http://www.twitter.com/audriga <http://www.twitter.com/audriga>
> 
> --------------------------------------------------------------------------
> audriga GmbH |  Alter Schlachthof 57  | 76137 Karlsruhe
> Sitz der Gesellschaft: Karlsruhe - Amtsgericht Mannheim - HRB 713034
> Geschäftsführer: Dr. Frank Dengler, Dr.-Ing. Hans-Jörg Happel
> --------------------------------------------------------------------------
> On 08/07/2022 10:22, Julien Schneider wrote:
>> Hi all,
>> 
>> Thanks a lot for your answers, it really helps.
>> By re-reading the RFC at the light of your responses, I think I'm starting to understand.
>> 
>> Section 3.4.2 starts with:
>>    The SCIM protocol defines a standard set of query parameters that can
>>    be used to filter, sort, and paginate to return zero or more
>>    resources in a query response.  Queries MAY be made against a single
>>    resource or a resource type endpoint (e.g., "/Users"), or the service
>>    provider Base URI.
>> 
>> Here queries are defined only by filtering, sorting and pagination. Attributes are not part of it, which suggests that their usage do not fall under the "query" definition.
>> Now, one could wonder how would a filter/sort/pagination make sense on a single resource. I (personally) cannot see any use case. Maybe that should be clarified in the RFC, or even removed.
>> 
>> What is very confusing is that, although not listed in the query section "introduction", attributes have their own sub-section 3.4.2.5 under the query section. I guess (?) that was done to explicitly specify that attributes can be used on queries as well (and not only on basic operations). In that case, it would maybe be better to either:
>> - explicitly specify in the 3.4.2.5 subsection that attributes can be used with/on queries, but do not define a query by themselves
>> - remove the 3.4.2.5 subsection and (more explicitly) specify in the 3.9 section that attributes can be used on queries
>> 
>> Anyway, I think things are clearer now? Do you agree with those affirmations?
>> - attributes do not imply a query but can be used on queries
>> - only filtering, sorting and pagination define a query
>> - GET /Users/abcd?attributes=userName is not a query and do not require a ListResponse response
>> - GET /Users/abcd?filter=userName eq "Blah" is a query (which make little sense, but still) and requires a ListResponse response
>> - GET /Users?filter=userName eq "Blah" is a query and requires a ListResponse response
>> 
>> Thanks a lot again!
>> Julien Schneider
>> Tel: [phone]
>> Fax: [phone]
>> 
>> http://www.audriga.com <http://www.audriga.com/> | http://www.twitter.com/audriga <http://www.twitter.com/audriga>
>> 
>> --------------------------------------------------------------------------
>> audriga GmbH |  Alter Schlachthof 57  | 76137 Karlsruhe
>> Sitz der Gesellschaft: Karlsruhe - Amtsgericht Mannheim - HRB 713034
>> Geschäftsführer: Dr. Frank Dengler, Dr.-Ing. Hans-Jörg Happel
>> --------------------------------------------------------------------------
>> On 07/07/2022 20:42, Phillip Hunt wrote:
>>> In practical terms, only the presence of a filter defines a query that mandates a ListResponse.  The other time a ListResponse is needed is a GET at the root level (which many don’t permit) or at the Resource folder level.
>>> 
>>> Attributes/excludedAttributes are listed under both 3.4.2 and 3.9.  In practical terms, the presence of the attributes/ecludedattributes qualifiers does not indicate a ListResponse.  Attributes are simply permitted on any SCIM operation. POST, GET, Query, etc.  A simple GET with Attributes does NOT require ListResponse (unless a filter is present) as per Section 3.9 and shown in the example.
>>> 
>>> Phillip Hunt
>>> @independentid
>>> [e-mail] <mailto:[e-mail]>
>>> 
>>> 
>>> 
>>> 
>>>> On Jul 7, 2022, at 11:04 AM, Danny Zollner <[e-mail] <mailto:[e-mail]>> wrote:
>>>> 
>>>> Hi Phil,
>>>>  
>>>> I’m attempting to clarify what you’ve said here for both my own knowledge and others. Anything under a subsection of RFC 7644 3.4.2 counts as a query. Query endpoints can be /Users/{id}, /Users, /Groups, and although not explicitly listed - /AnyOtherResource/{id} as well.
>>>>  
>>>> Features that fall under the label of queries and therefore require a ListResponse type response would be:
>>>>  
>>>> Filtering – i.e.: GET /x?filter=userName eq “blah”
>>>> Sorting
>>>> Pagination
>>>> Attributes – both ?attributes= and ?excludedAttributes=
>>>>  
>>>> Looking at 3.4.2.1, where the query endpoints are listed, the following parts of 3.4.2.x make sense and what you’re saying is apparent. From my own observations, I think there may be hundreds of implementations out there that implement ?attributes and ?excludedAttributes without a ListResponse wrapper when the query URL is a known resource such as GET /Groups/456?excludedAttributes=members.
>>>>  
>>>> I think we’re still a while away from attempting major changes to the protocol and schema RFCs, but in the future I’d be interested in having a discussion on adding an exception to the ListResponse requirement for querying known resources. The attributes queries are a bit different from the others. For filter, sort and paginate you can only really use them against the root of the server or the root of a resource (i.e.: /Users) as I understand it. I can’t think of a scenario where you’d do a filter like GET /Users/123?filter=userName eq “x” – is there a use case there that I’m missing? Similarly, I can’t think of scenarios where you’d sort or paginate results for GET /Users/123.
>>>>  
>>>> Thanks,
>>>>  
>>>> Danny
>>>>  
>>>> From: Phillip Hunt <[e-mail] <mailto:[e-mail]>> 
>>>> Sent: Thursday, July 7, 2022 12:36 PM
>>>> To: Julien Schneider <[e-mail] <mailto:[e-mail]>>
>>>> Cc: Danny Zollner <[e-mail] <mailto:[e-mail]>>; [e-mail] <mailto:[e-mail]>
>>>> Subject: Re: [scim] [EXTERNAL] Query on a specific known resource
>>>>  
>>>> Julien,
>>>>  
>>>> You are not wrong.  All queries regardless of path MUST have a ListResponse.
>>>>  
>>>> It may seem logical to make the jump to just returning a single resource for a query that can only return a single result, but this is not permitted in the RFC. From a protocol point of view, allowing skipping ListResponse makes the protocol more complex because it creates “exceptions” which have to be handled. For example, what happens if no filter match etc.
>>>>  
>>>> For the group:  For historical information, the issue is actually more complex than it seems.   For SCIM queries SCIM’s profile of HTTP overrides both HTTP GET (retrieves a resource) and HTTP POST (create a resource) to perform a search function. In the case of GET, url based filters cause privacy concerns because of the leak of confidential information in URLs. The GET method does not allow request bodies. Because of thies, SCIM also supports http POST queries. The HTTP definition suggests creation a resource.  Dual purposing these methods for search queries created necessary complexity. One of the basic rules of thumb is that whenever a “filter” shows up in a request, the request becomes a SCIM Query which mandates a ListResponse.
>>>>  
>>>> I discussed this at length with authors of the HTTP specifications and worked with Julian Reschke (co-author of HTTP) to submit a proposal for a HTTP SEARCH method.  This would unburden GET and POST methods and simplify protocol overall.  Julian indicated to me at the time of finalizing SCIM (around IETF93), that this issue comes up frequently for him as one of the HTTP authors.  In the end, the HTTPbis WG chose not to create a new method because there is too much water under the bridge.  See: https://httpwg.org/wg-materials/ietf93/ietf-93-httpbis-search.pdf <https://nam06.safelinks.protection.outlook.com/?url=https%3A%2F%2Fhttpwg.org%2Fwg-materials%2Fietf93%2Fietf-93-httpbis-search.pdf&data=05%7C01%7CDanny.Zollner%40microsoft.com%7C61a3ef6cefda487b73ef08da603f27f4%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C637928122539415111%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000%7C%7C%7C&sdata=ahU2tOHPGFEjoaXkz16HWA6W50waT9m5kcYXXIIpRzA%3D&reserved=0>
>>>>  
>>>> Phillip Hunt
>>>> @independentid
>>>> [e-mail] <mailto:[e-mail]>
>>>>  
>>>>  
>>>> 
>>>> 
>>>> 
>>>> On Jul 7, 2022, at 12:57 AM, Julien Schneider <[e-mail] <mailto:[e-mail]>> wrote:
>>>>  
>>>> Hi Danny, hi Phillip, hi everyone,
>>>> 
>>>> Thanks for your answers. I think the confusing part (for me at least) is "Queries MAY be made against a single resource or a resource type endpoint ......." at the beginning of RFC7644 section 3.4.2., followed by "Responses MUST be identified using the following URI: "urn:ietf:params:scim:api:messages:2.0:ListResponse" ".
>>>> 
>>>> My interpretation is that "GET /Users/2819c223-7f76-453a-919d-413861904646?attributes=userName" is a query against a single resource, and should then have a "ListResponse" response? Where am I wrong here?
>>>> 
>>>> Thanks
>>>> 
>>>> 
>>>> Julien Schneider
>>>> Tel: [phone]
>>>> Fax: [phone]
>>>>  
>>>> http://www.audriga.com <https://nam06.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.audriga.com%2F&data=05%7C01%7CDanny.Zollner%40microsoft.com%7C61a3ef6cefda487b73ef08da603f27f4%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C637928122539415111%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000%7C%7C%7C&sdata=ugCOwiY5EXLH8EfrAacwK9%2FDCGnhMp3Sh7L%2Fo8WQNOA%3D&reserved=0> | http://www.twitter.com/audriga <https://nam06.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.twitter.com%2Faudriga&data=05%7C01%7CDanny.Zollner%40microsoft.com%7C61a3ef6cefda487b73ef08da603f27f4%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C637928122539415111%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000%7C%7C%7C&sdata=e6nmZrdOJiUBA%2F31N1TwKLeeQTuwCVla%2FE4UuOds7AU%3D&reserved=0>
>>>>  
>>>> --------------------------------------------------------------------------
>>>> audriga GmbH |  Alter Schlachthof 57  | 76137 Karlsruhe
>>>> Sitz der Gesellschaft: Karlsruhe - Amtsgericht Mannheim - HRB 713034
>>>> Geschäftsführer: Dr. Frank Dengler, Dr.-Ing. Hans-Jörg Happel
>>>> --------------------------------------------------------------------------
>>>> On 07/07/2022 04:51, Danny Zollner wrote:
>>>> Hi Julien,
>>>>  
>>>> RFC 7644 section 3.4.2 specifically is talking about queries. Retrieving or modifying known resources (i.e.: GET /Users/12345 ) does not require a ListResponse type response. A query of GET /Users?filter=displayname contains “contoso.com <https://nam06.safelinks.protection.outlook.com/?url=http%3A%2F%2Fcontoso.com%2F&data=05%7C01%7CDanny.Zollner%40microsoft.com%7C61a3ef6cefda487b73ef08da603f27f4%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C637928122539415111%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000%7C%7C%7C&sdata=%2B7DoMHkSX3An6CfG913vNi5zgP%2BAJY28kXSnR%2FgBr80%3D&reserved=0>” or GET /Users?attributes=userName would require a ListResponse type response, as it does not identify a specific resource in the query URL via ID value (i.e.: “12345” in the previous example). On the other hand, GET /Users/12345?attributes=userName does not require the ListResponse type response as it does identify a specific resource.
>>>>  
>>>> To explicitly answer the final question in your email – the expected response to GET /Users/2819c223-7f76-453a-919d-413861904646?attributes=userNamewould be the second example you provided.
>>>>  
>>>> Cheers,
>>>>  
>>>> Danny Zollner
>>>>  
>>>> From: scim <[e-mail]> <mailto:[e-mail]> On Behalf Of Julien Schneider
>>>> Sent: Wednesday, July 6, 2022 3:41 AM
>>>> To: [e-mail] <mailto:[e-mail]>
>>>> Subject: [EXTERNAL] [scim] Query on a specific known resource
>>>>  
>>>> 
>>>> Some people who received this message don't often get email from [e-mail] <mailto:[e-mail]>. Learn why this is important <https://aka.ms/LearnAboutSenderIdentification>	
>>>> Hi all,
>>>> 
>>>> I have a question about queries performed against a SCIM resource object (like "/Users/{id}").
>>>> 
>>>> The RFC (https://datatracker.ietf.org/doc/html/rfc7644#section-3.4.2 <https://nam06.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fhtml%2Frfc7644%23section-3.4.2&data=05%7C01%7CDanny.Zollner%40microsoft.com%7C61a3ef6cefda487b73ef08da603f27f4%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C637928122539415111%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000%7C%7C%7C&sdata=jbbn1Y43e0GFmChjRVbpMnTKZEqlB5TzLgiJmhz7ZRY%3D&reserved=0>) states: 
>>>> Responses MUST be identified using the following URI:
>>>>    "urn:ietf:params:scim:api:messages:2.0:ListResponse"
>>>> 
>>>> If I understand correctly, that means the "schemas" parameter of the response to those queries must be set to:
>>>> "schemas":["urn:ietf:params:scim:api:messages:2.0:ListResponse"]
>>>> 
>>>> While I understand how that applies to queries on a resource type endpoint (like "/Users") or on the SCIM server root, I don't understand how that applies to queries on a specific resource object.
>>>> If I understand correctly, queries on a specific resource object actually are quite identical to "retrieving a known resource" (https://datatracker.ietf.org/doc/html/rfc7644#section-3.4.1 <https://nam06.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fhtml%2Frfc7644%23section-3.4.1&data=05%7C01%7CDanny.Zollner%40microsoft.com%7C61a3ef6cefda487b73ef08da603f27f4%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C637928122539415111%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000%7C%7C%7C&sdata=wbOMKvqD9T7hP5OFtruAwY2r26XrKdncriLPSapazGQ%3D&reserved=0>) which are a GET on a specific resource, like:
>>>> GET /Users/2819c223-7f76-453a-919d-413861904646
>>>> Responses to those requests should have the "schemas" parameter set to the resource schema(s):
>>>> {
>>>>      "schemas":["urn:ietf:params:scim:schemas:core:2.0:User"],
>>>>      "id":"2819c223-7f76-453a-919d-413861904646",
>>>> ...
>>>> }
>>>> 
>>>> Now, how should the response to the following query should look like? And to what value should the "schemas" parameter of the response be set?
>>>> GET /Users/2819c223-7f76-453a-919d-413861904646?attributes=userName
>>>> 
>>>> Should it be:
>>>>    {
>>>>      "schemas":["urn:ietf:params:scim:api:messages:2.0:ListResponse"],
>>>>      "totalResults":1,
>>>>      "Resources":[
>>>>        {
>>>>          "id":"2819c223-7f76-453a-919d-413861904646",
>>>>          "userName":"bjensen"
>>>>        }
>>>>      ]
>>>>    }
>>>> 
>>>> Or something like:
>>>>    {
>>>>      "schemas":["urn:ietf:params:scim:schemas:core:2.0:User"],
>>>>      "id":"2819c223-7f76-453a-919d-413861904646",
>>>>      "meta":{
>>>>        "resourceType":"User",
>>>>        "created":"2011-08-01T18:29:49.793Z",
>>>>        "lastModified":"2011-08-01T18:29:49.793Z",
>>>>        "location":
>>>>    "https://example.com/v2/Users/2819c223-7f76-453a-919d-413861904646" <https://nam06.safelinks.protection.outlook.com/?url=https%3A%2F%2Fexample.com%2Fv2%2FUsers%2F2819c223-7f76-453a-919d-413861904646&data=05%7C01%7CDanny.Zollner%40microsoft.com%7C61a3ef6cefda487b73ef08da603f27f4%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C637928122539415111%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000%7C%7C%7C&sdata=1BDZsY%2B1DkxD3t2ZObBwf5UHRZlnoXdE9UFuuBZioS0%3D&reserved=0>,
>>>>        "version":"W\/\"f250dd84f0671c3\""
>>>>      },
>>>>      "userName":"bjensen"
>>>>    }
>>>> 
>>>> Thanks a lot in advance
>>>> 
>>>> 
>>>> -- 
>>>> Julien Schneider
>>>> Tel: [phone]
>>>> Fax: [phone]
>>>>  
>>>> http://www.audriga.com <https://nam06.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.audriga.com%2F&data=05%7C01%7CDanny.Zollner%40microsoft.com%7C61a3ef6cefda487b73ef08da603f27f4%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C637928122539415111%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000%7C%7C%7C&sdata=ugCOwiY5EXLH8EfrAacwK9%2FDCGnhMp3Sh7L%2Fo8WQNOA%3D&reserved=0> | http://www.twitter.com/audriga <https://nam06.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.twitter.com%2Faudriga&data=05%7C01%7CDanny.Zollner%40microsoft.com%7C61a3ef6cefda487b73ef08da603f27f4%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C637928122539415111%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000%7C%7C%7C&sdata=e6nmZrdOJiUBA%2F31N1TwKLeeQTuwCVla%2FE4UuOds7AU%3D&reserved=0>
>>>>  
>>>> --------------------------------------------------------------------------
>>>> audriga GmbH |  Alter Schlachthof 57  | 76137 Karlsruhe
>>>> Sitz der Gesellschaft: Karlsruhe - Amtsgericht Mannheim - HRB 713034
>>>> Geschäftsführer: Dr. Frank Dengler, Dr.-Ing. Hans-Jörg Happel
>>>> --------------------------------------------------------------------------
>>>> 
>>>> _______________________________________________
>>>> scim mailing list
>>>> [e-mail] <mailto:[e-mail]>
>>>> https://www.ietf.org/mailman/listinfo/scim <https://nam06.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.ietf.org%2Fmailman%2Flistinfo%2Fscim&data=05%7C01%7CDanny.Zollner%40microsoft.com%7C61a3ef6cefda487b73ef08da603f27f4%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C637928122539415111%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000%7C%7C%7C&sdata=lgmWfvOoa2uTk4zhZ2eC6txKydhzire0Hd85sQ4Ib5Y%3D&reserved=0>
>> 
>