[Openid-specs-heart] HEART Meeting notes 2016-05-16
Eve Maler
eve.maler at forgerock.com
Wed May 18 13:55:59 UTC 2016
I get this feeling that there's a huge potential here for "base-level" vs.
"meta-level" confusion, and we're (very slowly) picking these apart. By
"base-level", I mean the broad-strokes or overt interactions intended to
take place between the system entities we're working with, and by
"meta-level", I mean interstitial or subtle interactions. An overt
notification pattern seems like something that has been contemplated, but
there are other kinds of notifications Adrian is asking for.
This reminds me of the "many definitions for delegation" that we once again
discussed on the last call. I dug up the whiteboard diagram for this that I
made at IIW that Adrian kindly captured, and I attach it here in the hope
that it will be helpful both generally and for the notification discussion:
- *Delegation type 1:* Alice (RO) -> AS: Alice outsources the job of
authorization to her always-on authorization server, an "agent" acting on
her behalf. This is a "meta-level" function made possible by the UMA
protocol, but gets into "inside-baseball" levels of discussion to talk
about.
- This doesn't seem to have a natural notification analogue in
HL7/FHIR, and I'm struggling to imagine what one might look
like. Possibly
a failure of imagination on my part.
- *Delegation type 2:* RO Alice (RO) -> (RqP) Bob: Alice shares resource
access with someone. This is a "base-level" function of UMA -- that is,
it's basically the end-to-end sharing function UMA was designed to
accomplish.
- I'm sorry I didn't label this "type 1"! This seems to have a
natural notification analogue in HL7/FHIR already, as well as in the NZ
work etc. Lots of reasons why client app functionality might include
notifying an RqP.
- *Delegation type 3:* Data subject Alice -> Grantor someone else: Alice
doesn't directly control access to resources that are about her, but
delegates that function to someone else who interacts with an AS on her
behalf. This is extremely "meta-level". (Note that it corresponds most
closely to Adrian's Caregiver use case.)
- This doesn't seem to have an existing notification analogue in
HL7/FHIR, but I might be mistaken. I can certainly imagine why one might
want notifications between these two parties.
- Note that the UMA protocol doesn't currently track, bind, or
correlate the two parties' separate identities; this is, in fact, a
completely meta-concept at the business/legal level.
[image: Inline image 1]
*Eve Maler*ForgeRock Office of the CTO | VP Innovation & Emerging Technology
Cell +1 425.345.6756 | Skype: xmlgrrl | Twitter: @xmlgrrl
Check out the 2016 schedule <https://summits.forgerock.com> for *ForgeRock
Summits and UnSummits*!
On Wed, May 18, 2016 at 6:04 AM, Debbie Bucci <debbucci at gmail.com> wrote:
> Please keep it simple and terminology constant across threads ! If you
> expand too far too you will loose me. When I see JASON I think additional
> metadata tagging of information. Please don't introduce
>
> You asked about use of subscriptions to existing use cases.
>
>
> I did not suggest anything would be exchanged beyond a notification but I
> do (strongly) believe all service will log/audit
>
> At some point Alice (grantor?) may want need to know more. Doesn't
> grantor consent/authorize?
> On May 18, 2016 8:28 AM, "Adrian Gropper" <agropper at healthurl.com> wrote:
>
>> Debbie, you are not following my model. My model has no prior consent, no
>> consent policy representation in FHIR or any other way, and no consent
>> receipts. My model is purely based on the JASON model where the resource,
>> and therefore FHIR, are under separate jurisdiction from the authorization
>> policies. In my model, the Grantor's policies are never communicated on the
>> wire.
>>
>> Think of this as analogous to never communicating a private key of the
>> wire. My model has no shared secrets and policies are secret. The policies
>> associated with the FHIR resource are locked up in the UMA Authorization
>> Server and never leave. The only thing that leaves the AS is authorization
>> tokens.
>>
>> Therefore, in my model (an instance of the JASON model) there is no prior
>> consent because Bob gets his authorization, regardless of whether it's
>> one-time or reusable, when he presents to the Authorization Server. (I will
>> not use Authorization for Disclosure ever any more. It's to close and
>> confusing with Authorization of Disclosure in HIPAA.)
>>
>> Therefore, as far as the FHIR resource server is concerned, it only needs
>> to generate an empty notification each time Bob actually accesses a
>> resource and store the AuditEvent in case the Grantor wants to use FHIR to
>> get an Accounting of Disclosures at a later time.
>>
>> Adrian
>>
>> On Wednesday, May 18, 2016, Debbie Bucci <debbucci at gmail.com> wrote:
>>
>>> Authorization for disclosure. Adrian if I follow you... Alice grants Dr
>>> Bob access her data (receipt generated here). Dr Bob later accesses data
>>> for the first time. Resource [authorization ??] server logs the
>>> disclosure event and optionally pushes a message via a subscription to
>>> alert Alice that her data was accessed. At some time later Alice could
>>> query for both her receipts and disclosures to verify she granted access
>>> and it was Bob who accessed her data. Seems to me these may be core to
>>> use case in a US context but peripheral to the profiles for its unclear
>>> if there would need to be any scopes or resources sets to define to support
>>> it. Defining the content of the subscription message,disclosure or
>>> receipt probably out of scope.
>>> On May 17, 2016 10:55 PM, "Adrian Gropper" <agropper at healthurl.com>
>>> wrote:
>>>
>>>> I'm glad to hear they're in scope.
>>>>
>>>> Thing is, that notifications have two components: the endpoint and the
>>>> message format. As I recall, MVCR consent receipts are more about the
>>>> message format than the endpoint. I much prefer the FHIR Subscriptions
>>>> approach because it doesn't trigger a standards project related to the
>>>> format and therefore makes implementation much faster.
>>>>
>>>> All three notification types are perfectly compatible with a
>>>> human-readable format. In some cases, as the FHIR Subscription
>>>> standard points out, having a message content at all is an unacceptable and
>>>> unnecessary privacy liability. This applies to type (1) notifications where
>>>> any contact with the notification endpoint triggers a query to the FHIR
>>>> endpoint.
>>>>
>>>> Type (2) notifications need to be human-formatted by definition to meet
>>>> the intent of HIPAA and today's API Task Force recommendation. Nothing
>>>> would be served by trying to standardize the format as each FHIR endpoint
>>>> has a right to issue whatever warning it feels is appropriate to the
>>>> particular client and situation. Although there was interest in the
>>>> API discussion today in evolving a Model Privacy Notice related to this
>>>> Type (2) notification, that work, if and when it's available, is also meant
>>>> for human-readable format and it was deemed unlikely to cover all of the
>>>> warning situations. There are no privacy issues with Type (2) messages
>>>> because they need not contain any PHI. This makes them much more
>>>> user-friendly since they can be delivered via SMS and other insecure
>>>> channels.
>>>>
>>>> Type (3) notifications - A4D - are the only ones that might benefit
>>>> from a standard format. Even so, standard formatting is not expected - we
>>>> are perfectly happy to receive activity messages from banks and merchants
>>>> even though their format is all over the place. In addition, type (3)
>>>> messages do have privacy implications if they contain any PHI or even
>>>> indication of the metadata. Therefore, it would be preferable to make these
>>>> opaque or content-free and allow the recipient to query the FHIR server for
>>>> A4D if they choose.
>>>>
>>>> @Grahame: Does FHIR already specify an A4D resource?
>>>>
>>>> Is the discussion above an appropriate addition to the current use-case?
>>>>
>>>> Adrian
>>>>
>>>> On Tuesday, May 17, 2016, Debbie Bucci <debbucci at gmail.com> wrote:
>>>>
>>>>> Notifications were always in scope. One type of notification was in
>>>>> the form of a consent receipt. You may recall Sarah doing some research
>>>>> and comparing to ...was it HL7? The receipt could be use to acknowledge a
>>>>> number of state changes.
>>>>>
>>>>> I am very familiar with topics/queues and pub/sub from my SOA
>>>>> infrastructure days. A brief glance at subscriptions seem to be a similar
>>>>> pattern.
>>>>>
>>>>> Dont think its a scope issue. It could be a implementation choice in
>>>>> use cases to use a subscription to deliver a message that either
>>>>> acknowledges a receipt or triggers another process in the work flow
>>>>>
>>>>> I think that would cover your 3 examples listed below.
>>>>>
>>>>> On May 17, 2016 8:45 PM, "Adrian Gropper" <agropper at healthurl.com>
>>>>> wrote:
>>>>>
>>>>>> As mentioned in the other thread, there are three different kinds of
>>>>>> notification under HIPAA (and probably in the general case):
>>>>>> (1) a resource has changed - this is p53 of the NZ doc
>>>>>> (2) a client is being registered that requires specific affirmation
>>>>>> by the Grantor
>>>>>> (3) accounting for disclosures - the resource server has been
>>>>>> accessed by a client
>>>>>>
>>>>>> The FHIR spec lists 4 different standards for push notifications
>>>>>> http://hl7.org/fhir/subscription.html .
>>>>>>
>>>>>> I believe all three of these kinds of notification are in-scope for
>>>>>> HEART. The three endpoints do not all have to be the same, do not all have
>>>>>> to be UMA, and we may want to restrict some of the 4 options offered in the
>>>>>> FHIR spec.
>>>>>>
>>>>>> Can we add these three kinds of notification to the current use case
>>>>>> as proposed by Eve?
>>>>>>
>>>>>> Adrian
>>>>>>
>>>>>>
>>>>>> On Tuesday, May 17, 2016, Eve Maler <eve.maler at forgerock.com> wrote:
>>>>>>
>>>>>>> Supplementing with the link I referred to during the call:
>>>>>>>
>>>>>>> The UMA case study "Users Managing Delegated Access to Online
>>>>>>> Government Services" conducted by the New Zealand government available on
>>>>>>> the UMA wiki:
>>>>>>>
>>>>>>>
>>>>>>> http://kantarainitiative.org/confluence/display/uma/Case+Study%3A+Users+Managing+Delegated+Access+to+Online+Government+Services
>>>>>>> (PDF report downloadable here
>>>>>>> <http://kantarainitiative.org/confluence/download/attachments/76907066/NZ%20ProjectClosureReport_PoC%20Delegations-Final%20Version%20SIG%20redacted%20Rev1.pdf?api=v2>
>>>>>>> )
>>>>>>>
>>>>>>> ...documents in Sec 8.4.3.3.3 (p. 46) a desire for "notifications"
>>>>>>> that can be achieved by sending email or by other means, with a discussion
>>>>>>> in Sec 8.4.9 (p. 53) about sending push notifications to an UMA requesting
>>>>>>> party through a mobile app from a resource owner's smart device. The case
>>>>>>> study didn't, I believe, use FHIR.
>>>>>>>
>>>>>>>
>>>>>>> *Eve Maler*ForgeRock Office of the CTO | VP Innovation & Emerging
>>>>>>> Technology
>>>>>>> Cell +1 425.345.6756 | Skype: xmlgrrl | Twitter: @xmlgrrl
>>>>>>> Check out the 2016 schedule <https://summits.forgerock.com> for *ForgeRock
>>>>>>> Summits and UnSummits*!
>>>>>>>
>>>>>>> On Tue, May 17, 2016 at 7:04 AM, Debbie Bucci <debbucci at gmail.com>
>>>>>>> wrote:
>>>>>>>
>>>>>>>> Adrian was given the floor to discuss his concerns that he has
>>>>>>>> labeled for point of reference as “Independent Decision Support at the
>>>>>>>> Point of Care for Physicians and Patients"
>>>>>>>>
>>>>>>>> Adrian indicated there were 4 primary functions needed. Many in
>>>>>>>> the group translated it in different ways but Debbie’s version was:
>>>>>>>>
>>>>>>>> -
>>>>>>>>
>>>>>>>> Release of Information
>>>>>>>> -
>>>>>>>>
>>>>>>>> Notification of Disclosures
>>>>>>>> -
>>>>>>>>
>>>>>>>> Subscribe to changes
>>>>>>>> -
>>>>>>>>
>>>>>>>> Automated patient right of access.
>>>>>>>>
>>>>>>>> Eve translated the bullets to both technical and policy work out of
>>>>>>>> the UMA workgroup and pointed to relevant/related publication developed in
>>>>>>>> close coordination with NZ.
>>>>>>>>
>>>>>>>> Josh was concerned that Adrian perhaps was proposing an entirely
>>>>>>>> new standard. John and others were trying to get clarification
>>>>>>>> around what Adrian was referring to as Data Blocking.
>>>>>>>>
>>>>>>>> After an hour long of healthy discussion – Josh asked if there were
>>>>>>>> any procedure on how to document to handle such a discussion within the
>>>>>>>> group Adrian was concerned about HEART’s
>>>>>>>> functionality/relevance in relation to his concerns. Hope the
>>>>>>>> following points brought up in the meeting addresses these concerns:
>>>>>>>>
>>>>>>>> -
>>>>>>>>
>>>>>>>> HEART’s function is not to create standards but to create
>>>>>>>> profiles of existing recognized standards to encourage interoperable
>>>>>>>> implementations of these standards. Please see :
>>>>>>>> http://openid.net/wg/heart/charter/
>>>>>>>> -
>>>>>>>>
>>>>>>>> All members are encouraged to bring use cases to the group to
>>>>>>>> consider. The functional use case is broken down into
>>>>>>>> technical functionalities that are flagged as core or peripheral to the
>>>>>>>> standards being considered under HEART.
>>>>>>>> -
>>>>>>>>
>>>>>>>> We acknowledge that with both the Standards and profiles, there
>>>>>>>> may be implementation decisions made to bridge gaps where existing
>>>>>>>> functionality does not exist. We strongly encourage those
>>>>>>>> implementers to officially log the gaps and solutions they made back with
>>>>>>>> the relevant Standards Bodies in order to give feedback to future
>>>>>>>> iterations of the Standards.
>>>>>>>>
>>>>>>>> Standards development is an iterative process and versioning
>>>>>>>> control of standards and specs are an implementation reality we have to
>>>>>>>> deal with. Changes are made incrementally
>>>>>>>>
>>>>>>>> We concluded the discussion by asking Adrian to develop a
>>>>>>>> functional use case to bring to the group for consideration. Subsequently
>>>>>>>> there has been a recap discussion posted prior to the notes that we will
>>>>>>>> append to these meeting notes to round out context to the discussion
>>>>>>>>
>>>>>>>> Next week’s meeting (May 23rd) will focus on the comment submitted
>>>>>>>> by Mike Jones re: suggested changes to the HEART profile. Additionally,
>>>>>>>> if permissible, we will revisit and consider the Argonaut report and their
>>>>>>>> recommendations as part of this review. All members interested
>>>>>>>> that may have input to the impending updates, please make it a point to
>>>>>>>> attend.
>>>>>>>>
>>>>>>>> * AI:* Justin is developing a step by step analysis for us to
>>>>>>>> consider.
>>>>>>>>
>>>>>>>> Will cancel the next two meetings due to conflicts: May 30th (US
>>>>>>>> holiday) and June 5th (Cloud Identity Summit )
>>>>>>>> On June 12th We should be prepared to go back to the current use
>>>>>>>> case
>>>>>>>> https://bitbucket.org/openid/heart/wiki/Alice_Shares_with_Physicians_and_Others_UMA_FHIR
>>>>>>>> and stay with that use case until we complete our final draft
>>>>>>>> semantic profile.
>>>>>>>> Adrian's new use case is the next in the queue and welcome any
>>>>>>>> others that folk may want to submit.
>>>>>>>>
>>>>>>>>
>>>>>>>> _______________________________________________
>>>>>>>> Openid-specs-heart mailing list
>>>>>>>> Openid-specs-heart at lists.openid.net
>>>>>>>> http://lists.openid.net/mailman/listinfo/openid-specs-heart
>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>> --
>>>>>>
>>>>>> Adrian Gropper MD
>>>>>>
>>>>>> PROTECT YOUR FUTURE - RESTORE Health Privacy!
>>>>>> HELP us fight for the right to control personal health data.
>>>>>> DONATE: http://patientprivacyrights.org/donate-2/
>>>>>>
>>>>>>
>>>>
>>>> --
>>>>
>>>> Adrian Gropper MD
>>>>
>>>> PROTECT YOUR FUTURE - RESTORE Health Privacy!
>>>> HELP us fight for the right to control personal health data.
>>>> DONATE: http://patientprivacyrights.org/donate-2/
>>>>
>>>>
>>
>> --
>>
>> Adrian Gropper MD
>>
>> PROTECT YOUR FUTURE - RESTORE Health Privacy!
>> HELP us fight for the right to control personal health data.
>> DONATE: http://patientprivacyrights.org/donate-2/
>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openid.net/pipermail/openid-specs-heart/attachments/20160518/844888cf/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 206743 bytes
Desc: not available
URL: <http://lists.openid.net/pipermail/openid-specs-heart/attachments/20160518/844888cf/attachment-0001.png>
More information about the Openid-specs-heart
mailing list