First implementation

[abulte]

Thoughts for an implementation with the v2 version of croquemort (webhook).

Workflow

1. uData dispatches a signal when a resource URL is created or modified
2. udata-croquemort extension hooks to this signal
3. udata-croquemort sends a check request to croquemort (POST request) with a frequency (scheduled check) and a callback url (webhook)
4. The callback url is a udata-croquemort view that feeds the extra of the udata resource with the check results — the view will be called (and the extras refreshed) regularly since the check is scheduled
5. The frontend tests for the extras value of the resources and displays a status accordingly (not checked yet, OK at date XXX, KO at date XXX)

Challenges

• Authentication : how do we ensure croquemort is the one calling the udata-croquemort callback url? We obviously do no want anyone to be able to set the check results on uData resources.
• Match a check result and a resource: we probably need Webhook: add support for reference id etalab/croquemort#34 to be able to locate for which resource a check has been made when the webhook returns. Or use a join model url hash in croquemort <-> resource id in uData in udata-croquemort?
• resource URL modification case: when a resource URL has been modified, we probably need to delete the scheduled check on croquemort of the old URL.

[noirbizarre]

I think prior to that, we need to:

• extract all croquemort related code here
• abstract the link checking part in udata the same way it's done for tracker and udata-piwik

[davidbgk]

In ❤️ with point 4!

About challenges:

1. auth: I don't think it's a problem for now. Let's make it work first.
2. match: maybe reuse the underused notion of groups on the croquemort side or add something in this vein?
3. old sheduled checks: add an entry point in croquemort to allow that?