Modifications to events, especially to an instance of a recurring event, or an update to a modified instance can all seem quite complex. It’s not that bad once you understand the various bits that you need to provide for the receiving application to make sense of it. See the RFC5545 spec for details. This tools.ietf.org version or icalendar.org or the kanzaki version.
- UID – must have and must be unique
- SEQUENCE – numerically update to indicate which is the latest version
- RECURRENCE-ID – if modifying a single instance with in a recurring stream. Note that for a UID and SEQUENCE pair, the “RECURRENCE-ID” value for a recurrence instance is fixed. The event date may change but the “RECURRENCE-ID” does not. It must always be the original date as generated by the RRULE.
- LAST-MODIFIED – Probably a good idea top update this too
- if a user subscribes to an ICS URL, then the receiving application will check for updates on a regular basis at a frequency determined by it. It will be seen as a separate calendar. It will use the SEQUENCE numbers to determine the latest event details. There may be a delay in an update being made on the originating system and it being reflected in the receiving system. The only reliable way to be sure that users will have updates seen in their calendars is if you can get them to subscribe to the url (or ‘import’ the url, not the file)
- if an ics file is imported (eg: when event details are sent as an ics file attachment, not as a url, or if the user downloads the ics file from a url and then imports it), then it will be added as events into the calendar chosen by the user. Updates may overwrite IF the user re-imports the events AND the application is clever enough to match up the UID’s (and RECURRENCE-IDs if it’s part of a recurring set). It should then also use the SEQUENCE to determine if the details are actually the latest. There is no way on devices to force a user to load an updated ics file sent to them.
Modifications to recurring rules or rdates and modifications to instances of those recurring rules (themselves already a modification.)
Assume we have a recurring event. It starts with a SEQUENCE: 0
DTSTART;VALUE=DATE:20140421 RRULE:FREQ=MONTHLY;INTERVAL=1;BYDAY=3MO UID:1f4irrb5uh6q3cdfsnq15vd00g_R20140421@google.com SEQUENCE:0
This generates an infinite set of mondays, including the following
- 19th May
- 16th June
- 21 July
- 18 August
Now, one of the instances gets modified. (This could be just the description or the actual date.) A RECURRENCE-ID VEVENT is created for each instance modification and the SEQUENCE NO is updated. The recurrence-id matches the date in the recurring rule that it applies to. The combination of UID, RECURRENCE-ID and SEQUENCE helps us identify the instance that is being updated.
For a given pair of "UID" and "SEQUENCE" property values, the "RECURRENCE-ID" value for a recurrence instance is fixed.
DTSTART;VALUE=DATE:20140721 (ie no date change in this example) DTSTAMP:20140720T064853Z UID:1f4irrb5uh6q3cdfsnq15vd00g_R20140421@google.com RECURRENCE-ID;VALUE=DATE:20140721
Now assume the original RRULE is modified too, perhaps just the description or shock horror, the recurring date definition? It’s sequence-no is updated too BUT now what should happen to the RECURRENCE-ID instance? Should it be updated to indicate it is still valid? IE we now have:
- generated date 21 July 2014 with UID and SEQUENCE = 3 and
- RECURRENCE-ID;VALUE=DATE:20140721 with matching UID and SEQUENCE = 3
It appears that perhaps SEQUENCE numbers can increment separately, so one could conceivably receive a file with a RRULE with an equal to or greater sequence number than a RECURRENCE-ID. From this discussion, it is really not clear what the supplying calendar application will do:
Once UID 1/SEQUENCE 0 hits UID 1/SEQUENCE 1 (which would mean something significant about the set of instances has changed) all bets are off in regards to RECURRENCE-ID values that previously existed.
If the UID/RECURRENCE-ID/SEQUENCE matches a generated date from the rrule with same SEQUENCE (ie both have been modified), then what? One cannot just use eg: LAST-MODIFIED:20140620T200327Z. Who is to say that a mod of the RRULE was intended to overwrite the RECURRENCE instance? Quite possibly not.
Some lengthy related discussions here and in other parts of the thread:
- http://www.imc.org/ietf-calendar/mail-archive/msg05681.html (recurrence-ids fixed ?)
- http://www.imc.org/ietf-calendar/mail-archive/msg05696.html (Independent Sequence Numbers)
Complications that urge one to keep events as simple as possible:
Searching the web seems to indicate some folks believe RECURRENCE-IDS should not change. So should we assume that if the RECURRENCE-ID vevent is still in the file then it is still valid and should be included. (What does this mean for syncing and deletions? IE How does one ‘delete’ an existing event. Or should one always delete events from an ics file and reload ONLY what’s in the ics file?)