Loro's rich text CRDT

1. Fork the text

2. Submit proposal

3. Review

4. Merge/Cancel

EDIT: To slightly expand on this - there are many reasons for this intuition - the main, IMO, is that people like to work on text privately before showing it to people. Also, the mental fear of your text interrupted by someone else. There might be even more reasons.

> We interviewed eight people who regularly collaborate professionally on documents such as news articles, and several told us that they found real-time collaboration a stressful experience: they felt performative, self-conscious of others witnessing their messy work-in-progress, or irritated when a collaborator acted on suggestions before the editing pass was complete. When doing creative work, they preferred to have space to ideate and experiment in private, sharing their progress only when they are ready to do so.

> With asynchronous collaboration, this is possible: a user may work in isolation on their own copy of a document for a while, without seeing other users’ real-time updates; sometime later, when they are ready to share their work, they can choose to merge it with their collaborators’ edits. Several such copies may exist side-by-side, and some might never be merged (e.g. if the user changed their mind about a set of edits).

[1]: https://www.inkandswitch.com/peritext/

> The REG algorithm excels with its fast local update speeds and eliminate concerns about tombstone collection in CRDTs. For instance, if an operation has been synchronized across all endpoints, no new operations will occur concurrently with it, allowing it to be safely removed from the history.

If you remove these ops from history, does that remove the ability to time travel (per the home page "An antidote to regret, enabling historical edits traversal") or merge branches? How can we be sure an operation is synchronized?

If dropping these ops is necessary for speed/storage optimization but disables time-travel, is it possible to put the removed historical/tombstone ops into a "cold storage" that's optional and only loaded for time-travel use?

Yes. But squash can be supported.

> How can we be sure an operation is synchronized?

In Loro, we not only record the real-world timestamp efficiently, similar to Git, but also capture the DAG information. This approach ensures that if an operation (op) is particularly old, it will have many other ops depending on it. By utilizing both pieces of information, we can determine the operations that are likely synced across all peers. For peers like servers, it's feasible to preserve all operations. However, we can remove some operations in scenarios such as opening the document online for the first time.

> If dropping these ops is necessary for speed/storage optimization but disables time-travel, is it possible to put the removed historical/tombstone ops into a "cold storage" that's optional and only loaded for time-travel use?

Yes. This is not supported at the moment, but we hope to implement it before version 1.0.

> If you remove these ops from history, does that remove the ability to time travel

Yes it does. You also need the ops from history to be able to merge changes. You can only merge changes so long as you have the operations going back to the point at which the fork happened.

> is it possible to put the removed historical/tombstone ops into a "cold storage" that's optional and only loaded for time-travel use?

Absolutely. And this is very practically useful. For example, you could have a web page which loads the current state of a document (just a string. Unlike CRDTs, it needs no additional metadata!). Then if some merge happens while you have the document open, the browser could just fetch the operations from the server back as far as it needs to be able to merge. But in normal operation, none of the historical operations need to be loaded at all.

All this said, with text documents the overhead of just keeping the historical operations is pretty tiny anyway. In my testing using diamond types (same algorithm, different library), storing the entire set of historical operations usually increases the file size by less than 50% compared to just storing the final text string. Its much more efficient on disk than git, and more efficient than other CRDTs like automerge and Yjs. So I think most of the time its easier to just keep the history around and not worry about the complexity.