What the SpaceDB note says a replica keeps

What the note claims

The SpaceDB field note describes local-first data management: a CRDT-native database whose replicas live near the user, work offline, and converge without a data center. That is the claim this page will defend, and it will not paste the rest of the note. The dependency is what a replica still holds when the network drops, and what has to be true for two replicas to agree later. The note that states the claim is the SpaceDB field note.

What local-first already meant

The Ink and Switch essay on local-first software argues that the user's device should hold the data, and that a network is for collaboration rather than for the system of record. The SpaceDB note is one implementation claim sitting on that argument. The essay does not mention the crate. The crate note does not replace the essay. A content address, a different name for a byte string, is what a content address is.

What a CRDT paper requires

Read together, the filed pages are the record, and a later paraphrase would not replace them. The Shapiro survey of conflict-free replicated data types describes updates that merge deterministically. Replicas that have delivered the same set of updates reach the same state without electing a server to decide the merge. Offline work is allowed because the merge is a function of the updates, not of a connection. The desk is Distributed Cloud.

Why it matters on this desk

The dependency is the system of record. If the replica near the user is the record, a data center outage is an inconvenience to sync, not a loss of the file. If the data center is the record, the replica is a cache and the note's sentence is false. This page does not audit the crate. It files the sentence next to the two public documents that define the words. A partition's effect on a write is what a partition does to a write.

What this page will not import

A replica that only fills in while a region answers is a cache, and the note's sentence would be the wrong description of it. The Ink and Switch essay is the public argument that the device holds the record and the network is for collaboration. The Shapiro survey is the public account of updates that merge because of their type, not because a server picked a winner. The field note names a database that claims both. This page does not audit the crate, does not count users, and does not draw a region map. Other house pages that mention search or access-control tiers are not the sentence filed here. Canonical's websites can be down while a local replica is still the record, which is why those briefs are different dependencies. The check is the location of the record: near the user, or in a data center the note says it does not need. Feature lists that appear on other house pages, including search, encryption details, and access-control tiers, are not in the one-sentence claim this brief is using. Importing them would be a second article. The October 8 Canonical disruption is a different document, about websites rather than a replica, filed as what Canonical's October 8 statement says was down.

Where the documents split

The essay argues for local-first software in general. The survey defines a family of data types. The field note names a database that claims to be both. They do not, together, prove a benchmark, a user count, or a region where a replica is running. The overlap is the shape: data on the device, merge without a center. The proof of a particular install is not in this file. The front of the paper is Futureweb.

What to watch next

A measurement that shows two replicas diverging after the same updates would falsify the CRDT claim for this database. A note that moves the system of record to a hosted region would falsify the local-first sentence. Another paraphrase of the same field note would not. The essay and the survey stay the public meanings of the words.

What would change the record

If the field note drops the offline sentence, this brief's claim has to drop with it. If a later survey replaces the merge definition, the citation moves. The Canonical outage does not change either document. A reader can lose ubuntu.com for an afternoon and still hold a local replica, which is the point of filing them as different dependencies.

What stays on the replica

What stays specific is a field note that calls SpaceDB a local-first, CRDT-native database whose replicas live near the user, work offline, and converge without a data center. The Ink and Switch essay is the public argument that the device should be the system of record. The Shapiro survey is the public definition of a merge that does not need a central decider. This page does not paste the note, does not audit the code, and does not invent a deployment map. The dependency a reader can check is the location of the record: near the user, or in a data center. The note picks the first. The other two documents explain what that pick has to mean if the words are being used in their public sense. A cache that only works while a region answers is a different product, and it would need a different sentence.