Upcoming changes to iCalendar and vCard validation in sabre/dav 3.2.

We are currently working on sabre/dav 3.2. A first beta has been released. This release will include some changes to validating iCalendar and vCard. These changes might impact you if you use the Card- and CalDAV systems, or if you are a developer for a Cal/CardDAV client.

In the past we've been pretty lenient in terms of what kind of data you can send the server. Originally the idea for this was to follow Postel's law, e.g.:

Be conservative in what you send, be liberal in what you accept

I actually believe that Postel's law and following it is not a good idea and will actually over time cause networks to become less and less compliant.

So starting 3.2 we are dialing up the strictness of the server. The component inside of sabre/dav responsible for parsing iCalendar and vCard has had a validation system for a while. sabre/dav now uses that system when you PUT a new iCalendar object or vCard.

The system is not yet 100% complete, so it will not yet throw errors on every invalid object, but over time we will make this more strict by adding new validation rules.

How it works

One of the biggest areas in validation is that we're now checking for properties that are required, and how many instances of properties may appear.

For example, in iCalendar the PRODID is required to appear in the top-level iCalendar object.

In that particular case, the system is actually able to repair the incoming object. It will simply add a default PRODID.

A request for this might look as follows:

PUT /calendars/user/calendar/new-object.ics
Content-Type: text/calendar

BEGIN:VCALENDAR
BEGIN:VEVENT
UID:foo-bar
DTSTAMP:20160523T181200Z
DTSTART:20160524T090000Z
SUMMARY:Meeting
END:VEVENT
END:VCALENDAR

Since PRODID is missing, sabre/dav will do the folowing:

  1. It will automatically add a PRODID
  2. It will no longer send back an ETag header after the request.
  3. It adds an X-Sabre-Ew-Gross header. This header is an indicator something was not right in the request, along with a description for a developer what was wrong with it.

The reason we're not returning an ETag is because we have to make modifications to the object. Not returning an ETag pretty much tells the client: you must do a GET request after your PUT to find out the current state of the object and the correct ETag. Fortunately, most clients do this correctly.

It also works well as a small punishment to a client. They are forced to do an extra HTTP request, so there is an incentive to fix the bug.

When we can't do a repair

There are many cases where we can't guess what the developer's intent was. For example: every VEVENT must have a DTSTART. If it didn't appear, we can't really just make one up.

For situations like that, we now always emit HTTP error 415 Unsupported Media Type, along with our X-Sabre-Ew-Gross header.

Requesting strict handling

There are also cases where a client developer does not want the server to automatically repair the object, and instead always do a hard failure when the server deems an object invalid. This might be especially handy during development.

To tell the server to always do hard failures, you can simply include the standard Prefer: handling=strict HTTP header in your PUT requests.

There are likely going to be compatibility problems

Because we are becoming more strict, it is likely that there are CalDAV and CardDAV clients that stop working, because they were sending us invalid data.

To deal with this, we are doing the following:

  1. We're testing popular clients to see if they have problems.
  2. If we run into those problems, we try to contact the developer of the client to see if they are able to fix it.
  3. If they are not fixing the problem, or if we determine that there will be many users stuck on an old version of their client, we will try to see if we can add an 'automatic repair' rule to the validation system.

The automatic repair will account for most, if not all client bugs, but if we do run into an issue with a client that will not behave correctly if we don't return an ETag, and the developer is unwilling to fix the problem (or enough users can't upgrade) we will consider a client-specific workaround, preferably based on the User-Agent.

However, we can't test every client on the face of the planet. So we're hoping developers and users of clients will test sabre/dav 3.2 with their software and report issues they run into.

Because we completely assume that we will run into new compatibility issues, and we also completely assume that not every client will be tested when the 3.2 release hits the floor, we recommend users of sabre/dav to test CalDAV/CardDAV clients they support before upgrading, or waiting a few point releases so the dust can settle.

Contact us

Are you running into a compatibility issue? Contact us via our Github issue tracker.

Tags: dav, icalendar, vobject, vcard


Comments