Local-first software: You own your data, in spite of the cloud

The essay argues that cloud apps and old-fashioned local apps each capture only part of what users need. Cloud services make collaboration and cross-device access easy, while local software gives users ownership, offline use, longevity, privacy, and control. It proposes local-first software as a model that keeps the primary copy of data on the user’s device and uses servers only as support.

Reading notes

#

• Cloud apps make it easy to collaborate and access data from any device, but they depend on the provider’s server and limit user control over the data.
• Creative users often want to keep the files they make for future reference, portfolios, or archives, and the essay treats that as a matter of ownership.
• In cloud apps, access goes through the server, so if the service is unavailable or shuts down, the user can lose access to both the software and the data.
• Local apps store files on the user’s disk, which lets users back them up, archive them, edit them with other tools, and delete them without asking permission.
• Local-first software keeps the local copy as the primary one, while servers hold secondary copies to help with multi-device access and sync.
• One ideal is fast response without spinners, because local reads and writes avoid waiting for a server round-trip.
• Another ideal is multi-device use without trapping work on one device, with synchronization across laptop, tablet, and phone.
• Local-first software should work offline, including on mobile devices and in places with poor connectivity.
• It should support seamless collaboration with multiple people, including real-time editing and other workflows such as suggestions or pull-request-style review.
• It should support long-term access, so work can still be opened and modified even if the original company or service is gone.
• It should provide security and privacy by default, avoiding centralized databases that collect everyone’s data in one place.
• It should give users ultimate ownership and control, meaning they can copy, modify, back up, and manage their data freely.
• The essay compares several models, including files and email attachments, web apps, file-sync services like Dropbox, and Git/GitHub.
• Files and email give strong offline use, longevity, and control, but weak collaboration.
• Pure web apps score well on collaboration and multi-device access, but poorly on offline use, longevity, privacy, and control.
• File-sync services work well for local files and backups, but real-time collaboration is weak and conflicts are common.
• Git and GitHub come close to local-first for code, with offline use and user control, but they are weak for real-time collaboration and non-text file formats.
• The essay also compares app-development backends such as thin web apps, thick mobile clients, Firebase, CloudKit, Realm, and CouchDB.
• It finds that no existing storage layer fully satisfies the seven ideals.
• The proposed technical direction is CRDTs, which allow concurrent edits to merge automatically and can support real-time collaboration across devices.
• The prototypes Trellis, Pixelpusher, and PushPin are presented as experiments built with CRDTs.
• The findings say the technology works, offline work feels good, conflicts are less common than expected, and visualizing history is important.
• The essay also says URLs are a good way to share documents, while peer-to-peer systems still need better networking and usability.
• It concludes that local-first software is possible, but still needs better tools, research, and products before it can replace server-centered systems.