In this Notes article, I roughly jotted down notes in English from the content of the English session video/podcast. This article is intended to be read as a reference while watching the video. I hope this article helps you even a little when you watch the actual video. (Please feel free to contact me if there are any errors or typos! m(__)m)

This article is based on Firebase offline: What works, what doesn’t, and what you need to know (Firebase Summit 2019).

Should we care about offline anyway?

#

• Yes! Users may..

  • be in crowded place

  • go underground

• Firebase is not an offline-first platform, but offline-tolerant

What works

#

• It (mostly) works

  • Caching things locally

  • Exponential back-off and retry

Firebase Auth

#

• you only have the access to the user object

• you can’t login

Gotchas

#

• Login

• Auth Persistence

  • Tells how long you stay logged in

  • Trade-off between offline-experience and security

Cloud Firestore

#

How large is that cache, anyway?

#

• For mobile - 100MB

• For web - 40MB

• on-device cache is not indexed right now

  • unpack and scan every document you are searching

  • if your cache is too big or too many writes this can be quite slowly

• Maybe not pre-loading your cache too much is better than overloading cache data

• Adjust your cache size by configuring Firestore

Cloud Storage

#

• you can save large object to the cloud

• specifying a reference like a path to upload and download the content

All The Measurement Libraries

#

• On Android, all this uploading is done by Google Play Services

  • That is why if your app were crashed on Android, you can see that crashed data and upload it for you

  • if offline, exponential back-off retry

• Performance Monitoring

  • 10MB of data, shared across all apps (Android)

  • 10MB of data, per app (iOS)

  • Oldest data usually gets eliminated

• Analytics

  • About 100,000 events

  • Most recent data gets eliminated

• Crashlytics

  • 9 crashes(up to 150k each)

Conclusion

#

• If you’re offline, things mostly just work

• Designed for occasional offline moments, but not offline-first

  • if you need complete offline experience, maybe not the right tool

• Measurement tools don’t trust data that too old

  • Mostly 72 hours old?

Impressions / Summary / Key Takeaways

#

• On Android, Firebase Analytics and Firebase Crashlytics send logs via Google Play Services.
  • Therefore, they work even if the app task is killed.
• Firebase is not an offline-first service, but it anticipates offline states from regular use (subways or crowded places) and uses two methods for offline support:
  • Local cache
  • Exponential backoff and retry
• It might not be the right tool for apps where providing the same service offline as online is essential.

That’s all!