New Categories for Matrix Spec Changes

2020-04-29 — General — Andrew Morgan

On April 14th, the Spec Core Team conducted a long-overdue retrospective about the things that were working in the Matrix Spec Proposal process, and those that were not.

The most glaring item on the list was the sluggish pace that many Matrix Spec Changes (MSCs) take throughout the proposal process, as well as the general lack of activity from the Spec Core Team members on proposals that have not yet started a Final Comment Period.

We deeply apologize for the frustration this has likely caused many MSC authors, and want to shed some light on the reasoning behind it, and what we plan to do to prevent leaving authors in the dark about why there may be no Spec Core Team activity on their proposal.

Proposal Triaging

There are currently 136 open MSCs that have yet to undergo Final Comment Period (FCP), 75 of which are marked as proposal-in-review, and 20 that have a FCP proposed. Relative to the 65 MSCs that have ever been closed, this is a lot of outstanding ideas, features and maintenance changes.

The Spec Core Team itself is made up of 8 members, each of which have separate full-time jobs. All team members are well-placed to be on the team given their wide breadth of knowledge across the Matrix ecosystem, however the majority are some of the most busy pushing forward Matrix's reference implementations - without which, Matrix will unquestionably fail. This limits the amount of MSCs that the team can effectively work on at a given time.

The team understands that there are MSCs that provide incredibly useful features, such as support for LaTeX in messages or the ability to "knock" on rooms, and would undoubtedly like to see them land at some point.

But there is also a large backlog of MSCs that provide even more fundamental fixes and additions to the protocol that the team needs to prioritise. These include things like cross-signing devices, the communities rewrite and finally merging reactions and edits into the spec.

While we announce what MSCs we're focusing on during a given week during TWIM, it's not as clear which items we're looking to pull from the backlog next. To help tackle this, and to help keep us honest, we've begun putting each MSC into either "feature", "maintenance", or "core" buckets. This materialises in the form of github tags, which can be used to filter the list of MSCs like so: feature, maintenance, core. For a given timespan, we’ll pick a track and pull MSCs out of that category when possible. More information about MSC categories are now detailed on the proposals page.

As for the next 6 to 12 months, we plan to work on items from the “core” category. We need to get Matrix to a point where it can compete with other, proprietary chat protocols and items in "core" are decidedly the proposals that will take us the furthest in that direction. This doesn't mean we won't occasionally look at an MSC in a different category, but it will heavily influence our prioritisation.

Future

We'll try this approach out over the next few months and see how it goes. The next Spec Core Team retro will occur in the middle of May, where we will review the process once again.

For now, if you have any feedback please come and chat with us in #matrix-spec:matrix.org :)