Sync Data Best Practices (Needs update)
The new sync API uses protobufs to communicate with Chrome services, which is nice because protobufs were written to be robust against protocol changes (see protobuf docs for details). However, once you start syncing data, changing your protobuf format isn't completely painless; not only can sync users upgrade from one Chrome version to another, but they may have different Chrome versions running at the same time! Fortunately, there are some best practices to help make it easy:
- Avoid using "explicit version numbers" in your protobuf; instead, have your code test for the existence of a field to determine what to do. Protobufs were written precisely to avoid version-specific logic, and testing for fields is more robust.
- Adding a new field is the simplest case. Old clients will simply ignore the unknown new field, and when old clients send up new data, the sync backend preserves unknown fields so new clients can still use them.
- Removing an old field is also pretty simple. Simply stop populating the field. Old clients will continue to use it, and new clients will ignore it. Once the Chrome stable version moves past the last version to use the old field, (where "moves past" means something like "the number of users using a version of Chrome older than the stable version drops below 1%) then you can remove the field entirely via a server-side map-reduce.
- Avoid repurposing existing fields. Instead, add a new field for the new data and stop populating the old field, although continue to read it if the new field isn't present. Then, when the stable version of Chrome moves past the last version that creates old data, add code to migrate the old field to the new field (preferably on the server side, as adding substantial migration code might cause sync traffic spikes on version upgrades). Finally, once the stable version of Chrome becomes the first Chrome version with the migration code, you can remove data in the old field.