[Openid-specs-heart] Principles for selecting "Vanilla" OAuth vs. UMA
jricher at mit.edu
Wed Jul 1 19:50:37 UTC 2015
Of the set, I prefer option #2. That is to say, if we can solve it simply with OAuth, then let’s just do an OAuth version of it. If you’ve got a way to apply UMA to it, or it requires UMA, then sure, do so. But I think that we should simplify the alice-to-alice case to vanilla OAuth wherever we can.
Remember, this is for understanding specific use cases, and not for the work that HEART is doing overall. We’ll still produce profiles for OAuth and UMA (and the various bits and bobs that go with that). The question is when to apply which one to our use cases to understand what’s happening.
> On Jun 29, 2015, at 5:12 PM, Josh Mandel <Joshua.Mandel at childrens.harvard.edu> wrote:
> On today's call we discussed a use case where a patient can help connect her patient portal (a.k.a. her EHR account) account to an external PHR. This is a great, common use case that we know we could handle with either "vanilla" OAuth, or UMA, or both. Of course, software systems need to know, up front, whether they'll be talking vanilla OAuth or UMA -- because the wire protocols are different.
> The question: When HEART encounters a use case like this, by which principle(s) we should select vanilla OAuth vs. UMA? Some examples of principles (to stimulate discussion) might be:
> Example principle #1: "Do all the things"
> We should produce two profiles each time this kind of situation comes up: one describing how to do it with vanilla OAuth, and one describing how to do it with UMA. This provides maximum flexibility for implementers with different needs/contexts.
> Example principle #2: "KISS"
> Any time vanilla OAuth can handle a use case, we should use vanilla OAuth. Save UMA for when it's required. This provides a simpler environment with fewer moving parts and stronger out-of-the-box software library support.
> Example principle #3: "UMA everywhere"
> Use UMA across the board, and avoid vanilla OAuth. Since UMA handles a more general set of use cases, and there's value in consolidation, UMA should be the preferred option in all cases. This way, implementers only ever need to do one (very general) thing.
> (I've tried to state these examples neutrally, but I must admit bias in favor of #2. Does that come through?)
> Looking forward to discussion,
> Openid-specs-heart mailing list
> Openid-specs-heart at lists.openid.net
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Openid-specs-heart