MSC4140: Cancellable delayed events

[toger5]

Rendered

This could also supersede MSC2228 (by making it possible to send a redaction with the /send endpoint. This is the case as mentioned here)

Implementations

• Synapse
  • Support MSC4140: Delayed events (Futures) element-hq/synapse#17326
  • MSC4140: finalised delayed events, and more element-hq/synapse#19038
  • MSC4140: add dedicated endpoint for delayed events element-hq/synapse#19354
  • MSC4140: put delay_id in unsigned data for sender element-hq/synapse#19479
  • MSC4140: update error responses element-hq/synapse#19539
  • MSC4140: allow auth on management endpoints for delayed events element-hq/synapse#19794
  • MSC4140: rate limit management against delay_id element-hq/synapse#19830
  • MSC4140: support getting a single delayed event  element-hq/synapse#19926
• JS-SDK
  • Support MSC4140: Delayed events (Futures) matrix-js-sdk#4294
  • Delayed event management: split endpoints, no auth matrix-js-sdk#5066
  • MSC4140: add dedicated endpoint for delayed events matrix-js-sdk#5133

Known differences between current implementations and the proposal

• The implementations use the error code M_MAX_DELAY_EXCEEDED, HTTP 400, and a response property max_delay when the server refuses to schedule an event because the requested delay is too large. The proposal has since switched to the error code M_FORBIDDEN, HTTP 403, and publishes the maximum allowed delay in the m.delayed_events capability (so that clients can discover it ahead of making the request).
• The implementations use the existing endpoints /send and /state with a new query parameter delay for scheduling delayed events. The proposal has since switched to a dedicated endpoint PUT /_matrix/client/v3/rooms/{roomId}/delayed_event/{eventType}/{txnId} where the delay is included in the request body.
  • The work to update the implementations was started in MSC4140: add dedicated endpoint for delayed events element-hq/synapse#19354 and MSC4140: add dedicated endpoint for delayed events matrix-js-sdk#5133. To finish it, another /versions key would be required to manage the migration from the previous unstable implementation.
• The implementations use delay and max_delay rather than delay_ms and max_delay_ms.
• The implementations use running_since rather than scheduled_at.
• The implementations support the delegation feature by using delay_id-based authentication on the management endpoints, rather than expecting standard authentication with a user access token.

Implementations in Element Call via the Widget API

These are only informational and shouldn't be relevant for the proposal process. The MSC doesn't depend on widgets and widgets are themselves not part of the spec.

• Element Call running in SPA mode
  • Bump matrix-widget-api element-hq/element-call#2529
• Element Web running Element Call in embedded mode
  • Support sending delayed events (Futures) / MSC4157 matrix-widget-api#90
  • Split delayed event management methods matrix-widget-api#143
  • Bump matrix-widget-api element-hq/element-call#2529
• Element X running Element Call in embedded mode
  • Future Events: Add future endpoints to ruma. ruma/ruma#1845
  • widget-driver: Support for sending future events through the widget api. matrix-rust-sdk#3600
  • Bump matrix-widget-api element-hq/element-call#2529

---

SCT stuff:

MSC checklist

FCP tickyboxes

Designated reviewers:

• Security team (requested out of band)
• @anoadragon453 with Backend experience

[turt2live]

With the assumption that the alternative gets further information added into the MSC, this appears ready to go (though there's no comments on the M_FORBIDDEN thread 😇)

@mscbot fcp merge

[mscbot]

Team member @turt2live has proposed to merge this. The next step is review by the rest of the tagged people:

• @clokep
• @dbkr
• @uhoreg
• @turt2live
• @ara4n
• @anoadragon453
• @richvdh
• @tulir
• @erikjohnston
• @KitsuneRal

Once at least 75% of reviewers approve (and there are no outstanding concerns), this will enter its final comment period. If you spot a major issue that hasn't been raised at any point in this process, please speak up!

See this document for information about what commands tagged team members can give me.

[turt2live]

MSCs proposed for Final Comment Period (FCP) should meet the requirements outlined in the checklist prior to being accepted into the spec. This checklist is a bit long, but aims to reduce the number of follow-on MSCs after a feature lands.

SCT members: please check off things you check for, and raise a concern against FCP if the checklist is incomplete. If an item doesn't apply, prefer to check it rather than remove it. Unchecking items is encouraged where applicable.

MSC authors: feel free to ask in a thread on your MSC or in the#matrix-spec:matrix.org room for clarification of any of these points.

• Are appropriate implementation(s) specified in the MSC’s PR description?
• Are all MSCs that this MSC depends on already accepted?
• For each new endpoint that is introduced:
  • Have authentication requirements been specified?
  • Have rate-limiting requirements been specified?
  • Have guest access requirements been specified?
  • Are error responses specified?
    • Does each error case have a specified errcode (e.g. M_FORBIDDEN) and HTTP status code?
      • If a new errcode is introduced, is it clear that it is new?
• Will the MSC require a new room version, and if so, has that been made clear?
  • Is the reason for a new room version clearly stated? For example, modifying the set of redacted fields changes how event IDs are calculated, thus requiring a new room version.
• Are backwards-compatibility concerns appropriately addressed?
• Are the endpoint conventions honoured?
  • Do HTTP endpoints use_underscores_like_this?
  • Will the endpoint return unbounded data? If so, has pagination been considered?
  • If the endpoint utilises pagination, is it consistent with the appendices?
• An introduction exists and clearly outlines the problem being solved. Ideally, the first paragraph should be understandable by a non-technical audience.
• All outstanding threads are resolved
  • All feedback is incorporated into the proposal text itself, either as a fix or noted as an alternative
• While the exact sections do not need to be present, the details implied by the proposal template are covered. Namely:
  • Introduction
  • Proposal text
  • Potential issues
  • Alternatives
  • Dependencies
• Stable identifiers are used throughout the proposal, except for the unstable prefix section
  • Unstable prefixes consider the awkward accepted-but-not-merged state
  • Chosen unstable prefixes do not pollute any global namespace (use “org.matrix.mscXXXX”, not “org.matrix”).
• Changes have applicable Sign Off from all authors/editors/contributors
• There is a dedicated "Security Considerations" section which detail any possible attacks/vulnerabilities this proposal may introduce, even if this is "None.". See RFC3552 for things to think about, but in particular pay attention to the OWASP Top Ten.

[AndrewFerr]

To go even further with #4140 (comment), how about returning M_FORBIDDEN & HTTP 403 even for this case of requesting a delay longer than allowed?

This would prevent needing a special case error response for delayed events being entirely disallowed, as both that case & this one would then both give the same response.

It also looks like the spec uses M_INVALID_PARAM only for parameter values that are always invalid (like a malformed room alias or MXID), as opposed to values that are disallowed by server config that may change later.

[AndrewFerr]

Erm, actually, the special case would still be needed for when the maximum per-user amount of delayed events is 0, lest the response would be M_LIMIT_EXCEEDED & HTTP 429 for a limit that can never be satisfied.

But the point stands about M_FORBIDDEN & HTTP 403 potentially being more appropriate than M_INVALID_PARAM / HTTP 400.

[AndrewFerr]

In the interest of progressing with the Synapse implementation of MSC4140 error codes (PR), I've committed this change as 9447ee0.

I've also updated the PR description accordingly.

[Johennes]

But the point stands about M_FORBIDDEN & HTTP 403 potentially being more appropriate than M_INVALID_PARAM / HTTP 400.

This sounds sensible to me.

[gewitternacht]

There's a mismatch here with the wording in the Security considerations – Authentication section below, which says

As such, generated delay_ids MUST be cryptographically random such that they are difficult to guess.

1. "cryptographically random" vs. "cryptographically secure" – I think a more accurate and unambiguous wording in both places would be something like

   generated using a CSPRNG (Cryptographically Secure Pseudorandom Number Generator) and has sufficient entropy

2. MUST vs. SHOULD – I would change the SHOULD here to MUST, because that's what the authentication for these endpoints through knowledge of delay_id relies on.

[gewitternacht]

A bit later, "the cryptographic security of the delay_id" is mentioned again, which reads a bit weird to me. Alternatives that would sound better to me are for example

• the CSPRNG-generated delay_id, or
• the cryptographic randomness of the delay_id, or
• the unpredictability of the delay_id.

[Johennes]

A bit later, "the cryptographic security of the delay_id" is mentioned again, which reads a bit weird to me. Alternatives that would sound better to me are for example

* the CSPRNG-generated `delay_id`, or

* the cryptographic randomness of the `delay_id`, or

* the unpredictability of the `delay_id`.

This part was resolved in 5dac49c. The original comment above is still open though.

[gewitternacht]

An aspect I'm missing here is the fact that the POST /delayed_events/{delay_id}/{action} and GET /_matrix/client/v1/delayed_events/{delay_id} endpoints are including the delay_id, which is aptly described to "behave as a scoped access token". This risks a delay_id, i.e. an access token, getting leaked to various logs.

Can this risk be avoided / reduced by putting the delay_id in a header field or the request body instead? If there's technical reasons to not do that, this should at least be mentioned in the Security Considerations.

[Johennes]

I suspect the main reason is that in the path in makes for a very natural REST API shape.

I think putting it into the body would require changing GET /_matrix/client/v1/delayed_events/{delay_id} to POST. It's not as nice an API but would probably work.

As a header, we could leave the HTTP method unchanged. The API shape strikes me as equally odd but technically it should also be possible.

@AndrewFerr curious what you think? Have I missed any reasons why this needs to be in the path?

[AndrewFerr]

Yes, the main reason is to be able to use the delay_id as a typical "identifier" token.

Besides, there's limited effectiveness in trying to hide a delay_id by moving it into request headers / body / elsewhere, because it will nevertheless appear in client logs via /sync responses once its associated event gets sent, as per delay_id in unsigned event data.

[gewitternacht]

it will nevertheless appear in client logs via /sync responses once its associated event gets sent

But once the event gets sent, is there even a need to keep the delay_id secret any longer? Aren't all of the delay_id's access token capabilities void once the respective event is finalised?

See this part of the MSC:

If the target delayed event is already finalised with an outcome that conflicts with the action, i.e. if the action is send or restart and the delayed event has already been cancelled, or if the action is cancel and the delayed event has already been sent, the homeserver will respond with HTTP 409 and a standard error response with an errcode of M_UNKNOWN.

[AndrewFerr]

But once the event gets sent, is there even a need to keep the delay_id secret any longer? Aren't all of the delay_id's access token capabilities void once the respective event is finalised?

Good point, that's correct.

In the interest of unblocking this, I'll split off the delegation feature into its own MSC. For now, I've moved it to an alternative: cd878d2

---

With that out of the way, I'll say that I'm still hesitant to protect delay_ids as much as access tokens, for a few reasons:

• It would preclude future endpoints from being able to use delay_ids in ways that would "leak" them by design. A realistic possible addition is for /sync to include information about newly-scheduled delayed events (so that clients other than the one that scheduled an event would be notified of it, instead of having to manually hit the lookup endpoint to discover it).
• It may add friction against migrating to using OAuth 2.0 scopes for the management endpoints.

[gewitternacht]

With the delegation removed from this MSC, and the management endpoints being authenticated, I'm obviously fine with the security considerations in this MSC 👍

I'd be interested to see the additional delegation MSC when it's ready though – especially how it plans to deal with the authentication for the management endpoints prescribed in this MSC – and I think it would be good the mention the concerns form your comment about the handling of the delay_id there.