Re: [calsify] iCalendar Series draft

[Neil Jenkins <[e-mail]>]

On Sun, 25 Apr 2021, at 12:07, Michael Douglass wrote:
> *** Why standardize? ***

It was a bit more specific than that. I was concerned that to me the current proposal:
 * Doesn't seem todo enough to be really interoperable. It doesn't define who generates the events and when. Generation is an action rather than a data, so this really needs to be a CalDAV/JMAP Calendars extension, not just a data format extension.
 * Is overly complicated in the data representation. Reproducing all the recurrence properties as SRULE etc. seemed needlessly complicated to me. And using one of the events as both an event and the template for generation seemed a mix of concerns.
I like the idea though.  I think it could work like this:
 * We define a property (e.g. `seriesId`) for events generated in a series. Clients that understand this will know that these events are related. Clients that don't can happily ignore it.
 * Series are generated on a server. We define a CalDAV extension for servers to advertise that support this.
 * To configure series generation, you store a new component type (e.g. VTEMPLATE) into a calendar. This can contain the same properties as a VEVENT, so we don't need to define many new ones: just the ones about generation. Clients that don't support the extension will just ignore these components so won't interfere, but will still allow users to read and edit the generated instances perfectly (which are just regular VEVENTs).
 * The new properties we need would be just to configure how many instances to generate in advance (either by date, e.g. two weeks in advance, or by number, e.g. max 2 events from now). If you don't add any of these properties, you still have a template that could be manually used in clients that support templates.
 * The server would generate the events by doing the normal recurrence rule expansion, but strip the recurrence properties from the VEVENT output; they would be regular standalone events, with their own UID. Other than that all properties in the VTEMPLATE would be copied over.

I think that's most of it.

Neil.