Make clear the thread root is not in the thread

[andybalaam]

This was originally written up as MSC4037, but discussion there concluded that a spec PR was more appropriate.

Quoting @turt2live in that discussion:

MSC3771 (merged) originally introduced the threaded receipts behaviour into the spec in v1.4, but did not describe what actually constitutes what is "in a thread". Though, the implied wording of MSC3771 does not include the thread root in the thread for purposes of notifications. This is further backed up by Synapse being the primary cited implementation proof for MSC3771, which is backed up by this proposal's text in the above paragraph.

This PR modifies the spec to make clear that thread roots and their non-thread children are not "in the thread" for the purposes of receipt handling and unreadness.

• Synapse has the behaviour specified in this PR.

• matrix-js-sdk was recently fixed to match the behaviour required by this PR, as a bug fix for stuck notifications bugs resulting from inconsistency between Synapse and Element Web.

Signed-off-by: Andy Balaam [email protected]

Preview: https://pr1677--matrix-spec-previews.netlify.app

[andybalaam]

I believe this should have the spec-bug label but I don't have permission to add it.

[andybalaam]

I think the changelog entry should be .bugfix but the validator appears to reject items with that name?

[richvdh]

I think the changelog entry should be .bugfix but the validator appears to reject items with that name?

We use different suffixes in spec to other projects: https://github.com/matrix-org/matrix-spec/blob/main/CONTRIBUTING.rst#adding-to-the-changelog. clarification is right here, if the argument is just that the spec was "misleading".

As for labels: we don't normally bother with them on PRs for the spec.

[andybalaam]

@richvdh I was going by:

Likewise, corrections to the specification, to fix situations where, in practice, servers and clients behave differently to the specification, including anything with the spec-bug label.

From https://github.com/matrix-org/matrix-spec/blob/main/CONTRIBUTING.rst#other-changes

I guess this does fit under "clarification" from https://github.com/matrix-org/matrix-spec/blob/main/CONTRIBUTING.rst?rgh-link-date=2023-11-22T11%3A20%3A43Z#adding-to-the-changelog but if a .bugfix existed I think it would fit better, since we are correcting the spec to conform with implementations, because those implementations do what we meant to say first time.

[andybalaam]

@richvdh I was going by:

Likewise, corrections to the specification, to fix situations where, in practice, servers and clients behave differently to the specification, including anything with the spec-bug label.

Ah, but now I see that is referring to labels on issues, not PRs.

[turt2live]

Generally this looks good to me with a few comments.

For future PRs, please avoid re-flowing entire paragraphs as it makes it difficult to see what the changes actually are.

[turt2live]

We've now lost the definition of a "thread root" here.

[andybalaam]

Attempted a definition in 058e97e . I removed the original definition because it was not accurate (because it said the thread root was in the thread, which it is not).

[turt2live]

Thanks!