Matrix 1.7 has just been released! The last spec release was about 3 months ago, keeping us on track for regular quarterly releases. Unlike Matrix 1.6 though, today’s release is packed with plenty of features, some of which we’d like to call out here. Not all implementations will have support for these features yet though, and that’s okay (expected, even).
Adding support for a spec release can be a significant body of work. Instead of implementations having everything ready for spec release day, the idea is that they gain support over the next few months. If you’re able, please help those projects get v1.7’s features.
Today, we see 15 MSCs achieve their formally adopted status. All of them bring forward some much-needed features to Matrix, and a few highlights are below. Read on to the full changelog for a complete overview, and for a sneak peak at what the Spec Core Team (SCT) is planning to look at for v1.8 👀
The media repo has been largely unchanged since it was first released in 2015 as r0.0.0 - the only change being the addition of URL previews in r0.3.0 (released in 2017). Thankfully, the folks over at Beeper have been busy changing this situation!
MSC3860 and MSC2246 are both available as of today in the spec, improving how much bandwidth media repos need to offer and fixing a long-standing issue where uploading large files can be a challenge. MSC3860, specified here as
allow_redirect, allows clients to opt-in to HTTP redirects on downloads, avoiding slowdowns from the server having to proxy and move the media twice. It’s particularly useful if your server uses a CDN like S3 to host media - just tell clients to go grab media directly from S3 instead.
Meanwhile, MSC2246 (largely specified here) gives clients an ability to upload their media after sending the associated event. The MSC was originally designed with bridges in mind, where message order and delays are very perceivable problems for users, however clients like Element X are thinking about using the feature to improve local echo on uploads as well. We’re excited to see uploads finally be reliable, and non-blocking for the conversation flow!
History would indicate that we’ve got several years to go before the next major improvement to media, but we’re looking to change that: MSC3916: Authentication for media, MSC4016: Streaming E2EE file transfers, and MSC3870: Upload direct to URL are all fairly promising MSCs we’re planning to take a look at in the next year or so.
Reactions have actually been in use for quite some time already, but the MSC struggled to make it into the spec formally. With Matrix 1.7 though, MSC2677 is merged.
MSC2677 was the last part of the MSC1849 saga, with MSC2674 (relationships) and MSC2675 (server-side aggregation) landing in Matrix 1.3, and MSC2676 (edits) landing in Matrix 1.4 back in September 2022. There’s still some work to be done in this area though, and certainly some quality of life improvements asking to be written up as MSCs - watch this space for progress on those fronts.
One of the best parts about writing an MSC is eventually seeing it released in the spec, and these folks saw their MSCs reach this milestone for the first time. Congratulations everyone, and please keep them coming!
In no particular order:
Anyone can contribute MSCs, improving the whole of Matrix for everyone. If you have an idea (or bugfix), check out the guidelines and let us know if you run into any questions in #matrix-spec:matrix.org on Matrix - we’re always happy to help.
The SCT aims to cut a release in the middle of each calendar quarter, largely to avoid most conflicts with regional holidays, and since Matrix 1.1 the releases have been getting less and less organic. Instead of looking at MSCs whenever they’re raised, the SCT has been aiming to hear about which MSCs will need looking at. This is a very subtle difference, but one that dramatically changes how the quarterly releases are structured.
With each spec release, the SCT has slowly been working towards a place where the majority of MSC work is thought about in advance, and Matrix 1.8 is another milestone along this journey. Over the next couple weeks we’ll be working on a roadmap based on the MSCs that are raised to us in the SCT Office room on Matrix, starting with the ideas we raised ourselves earlier in the year.
As of writing, our current plan for Matrix 1.8 includes:
While this doesn’t represent a commitment to have these MSCs merged, these are the areas that the SCT is likely to be thinking about for the next 3ish months. If you have MSCs that might be possible to merge before roughly August, let us know in the SCT Office room (even if those MSCs aren’t on-theme with the above - we still want to hear about them!).
Critically, our release planning does not just include MSCs that are on track for being merged. We are also aiming to track that a given MSC needs technical review (for example), or that it might be receiving attention external to the SCT (such as implementation). We’d love to hear what you’re working on so we can start bringing those MSCs to the top of the list - let us know in the SCT Office room.
As is the theme for this section, if you have any questions about what the release process looks like (or where a given MSC currently sits in the roadmap), let us know in the SCT Office room.
There’s so many more things than what we covered in this blog post - flip through the changelog below for a full idea of what’s landed. Special thanks to MichaelKohler, zecakeh, and Cadair for contributing clarification & bug fix PRs in this release - we greatly appreciate it!
m.replace(edit) events, as per MSC3925. (#1440, #1525)
event_property_contains, as per MSC3758 and MSC3966. (#1464)
m.annotationrelations (reactions), as per MSC2677. (#1475, #1531)
.m.rule.is_room_mentionpush rules; and other notification behaviour, as per MSC3952. (#1508)
limiton a filter if one is not specified. (#1463)
state_defaultis 0 if there is no
m.room.power_levelsevent in a room. (#1479)
coalescepush rule actions, as per MSC3987. (#1501)
m.locationscheme by partially reverting f1f32d3. Contributed by @HarHarLinks. (#1507)
knock_restrictedjoin rule to the
age_tsfield from the reference hash calculation section. (#1536)
No significant changes.