open ended events should not require an end date

[gyst]

If you add an event, select a start date in the future (i.e. later than the default end date), then check 'Open End', you get a validation error "End date must be after start date.". This should not happen for open ended events. I'll submit a fix shortly.

[jackahl]

We just encountered this problem again using Plone 6 with Volto. Currently the Volto frontend also rejects saving when there is no end date set even though the event is open ended, but after resolving that in plone/volto#7995 or manually forging a POST/PATCH request the backend comes back with {'field': 'end', 'message': 'end is a required field. Setting it to null is not allowed.', 'error': 'ValidationError'}. So I think this should be looked at again.

P.S. Sorry for necromancy here ;)

[stevepiercy]

See also plone/volto#7993 (comment) for the RFC 5545 iCalendar specification for VEVENT standard and its interpretation.

[gyst]

P.S. Sorry for necromancy here ;)

Thanks for that. 2013, my god. Looking at my fix, it did not lift the requirement for having an end date, it just accepted an end date equal to the start date for open ended events. I see the plone/volto#7995 Volto fix uses the same approach. All of that is suboptimal. There should be an invariant that checks (in pseudocode): event.start < event.end OR (event.start AND event.open_end).. I.e. for open ended events, the requirement for an end date should be dropped.

The event schema has the end as a hard requirement though.

plone.app.event/src/plone/app/event/dx/behaviors.py

Line 96 in ffc29bb

required=True,

The field has a default value of one hour after the start

plone.app.event/src/plone/app/event/dx/behaviors.py

Line 97 in ffc29bb

defaultFactory=default_end,

I've left some comments on the Volto PR to suggest bringing the workaround there in sync with the actual backend behavior.

I'm not sure actually changing the schema is needed? I.e. replace the required with more complex invariant enforcement? The backend default behavior seems sane enough, especially because it results in a sane situation when you set open_ended=False in an update; in that case the provided default end date gets "unshadowed" again in a sane way. Note that the Volto PR should be updated to get to that sanity, it misses the one hour offset.

[stevepiercy]

See also my amended comment. I think the third attempt is the correct interpretation, and agrees more closely to @gyst's comments plone/volto#7993 (comment).

[stevepiercy]

Also I think we should make a PLIP of this use case. It's in many places, and centralizing the discussion would benefit everyone. I can do that tonight while you're all sleeping.