Migrate to tiptap v2

[juliusknorr]

Might need some additional evaluation to see what we actually need to change, especially with the custom marks/blocks and the collaboration/annotation plugins, but i think we should at some point consider an upgrade.

https://tiptap.dev/overview/upgrade-guide

[hanspagel]

Tiptap co-creator here, if someone needs help evaluating: Hit me up here or via email: [email protected] ✌️

[azul]

As a start i will summarize considerations from the last time we talked about this.

Why upgrade:

• stay up to date with our dependencies
• new, cleaner codebase
• benefit from new features and plugins
• longer term compatibility for things we develop ourselves
• Collaborative editing is fairly slow right now. Faster sync times of doc status would be nice.

Challanges with the Collaboration plugin

• We need a way to sync clients that can be deployed on a plain LAMP stack.
• WebRTC might not work because of lack of turn server in certain network setups
• We need to save the state of the document to the file storage
• Optionally improve performance with non LAMP addition

Possible approaches

Forward port the old collaboration plugin to tiptap2

There's the collaboration plugin for tiptap1 based on prosemirror - forward port it to tiptap2.
Following the suggestion in the tiptap2 migration guide
Would be the closest to what we have so far.

pros:

• no need to change server code
• should be pretty straight forward - it's actually very little code
• maybe reusible for other projects in a similar situation
• known to work on LAMP stack
• Would work in all deployments
• Could be used as a fallback

cons:

• cannot use tiptap2 plugin as is
• not possible to combine with webRTC
• no performance improvements
• no Y.js merges

PHP implementation of Y.js server side

pros:

• easy to integrate with nextcloud file storage
• Would work in all deployments

cons:

• websockets still seem to be quite a bit of a pain in php
• I did not find any php websocket code in nextcloud this far
• Might be cumbersome to write performant async code in php

Rust implementation of Y.js as part of the notify_push app

pros:

• Rust (type safety, high performance, async primitives)
• Rust implementation of Y.js is on the way
• push_notify is already written in rust and promoted as "high performance backend"

cons:

• Requires fallback for LAMP deployment
• Maybe harder to integrate with storage (but notify_push already connects to the storage)

Deploy Y.js server

pros:

• No need to (partly) develop a Y.js server.
• easy to deploy optionally and containerize (i guess)
• Can have POST requests to sync the datastorage

cons:

• Requires fallback for LAMP deployment
• We'd still need a way to store state to the files backend