FHIR Chat · service error · cds hooks

Stream: cds hooks

Topic: service error


view this post on Zulip David Hay (Jan 21 2018 at 02:57):

If there's an error in the service processing, is there a defined response? The 'patient-hello-world' just returns a string (if you send it a patient with no name), and Bryns service returns an html page with a screen dump...

view this post on Zulip Isaac Vetter (Jan 21 2018 at 03:03):

Hey David, there was some discussion around this question, here: https://github.com/cds-hooks/docs/issues/14

In summary, an appropriate HTTP status code seems appropriate, such that the CDS Hooks client knows not to show the response to a user and perhaps additional details if available.

view this post on Zulip David Hay (Jan 21 2018 at 03:09):

Thanks Isaac - but what format should the additional data be? I agree with Grahames last comment - it should be something that the client expects...

view this post on Zulip Isaac Vetter (Jan 21 2018 at 03:12):

Yeah, re-reading that thread, Sean and Kevin seem fairly set against OperationOutcome -- I think due to its additional overhead. I'd suggest that if the CDS Hooks service is able to explain its HTTP status code, it should do so in json, at the least.

view this post on Zulip Isaac Vetter (Jan 21 2018 at 03:14):

Hey @Kevin Shekleton , would it make sense to resolve #14 during the Jan WGM with community input? It looks like the right answer is currently disputed.

view this post on Zulip David Hay (Jan 21 2018 at 03:15):

btw - what's the agenda for the friday evening meeting?

view this post on Zulip Kevin Shekleton (Jan 22 2018 at 16:50):

We haven't set a firm agenda on the issues being discussed -- we'll get that landed this week and communicated

view this post on Zulip Kevin Shekleton (Jan 22 2018 at 16:51):

We can discuss #14 but we should ensure that we don't rehash and just have the same discussion all over again. The reason why that issue is still open is that we need to write up the guidance the community agreed upon and put that into our spec.

view this post on Zulip Kevin Shekleton (Jan 22 2018 at 16:54):

The core issue on #14 was if CDS Services should return error information that can be programmatically interpreted (more so than HTTP status codes). The broad consensus in the room at madrid (representing a diverse set of users) was that programmatic error handling beyond HTTP status codes wasn't valuable.

view this post on Zulip Kevin Shekleton (Jan 22 2018 at 16:55):

CDS Services that have an error should return a HTTP status code that is appropriate for the particular error.


Last updated: Apr 12 2022 at 19:14 UTC