matrix-js-sdk uses unstable /relations endpoint (and its unspecced original_event field)

[clokep]

The /relations endpoint does not return an original_event field, yet the matrix-js-sdk uses it.

We would like to remove support for this from Synapse in < 6 months. See matrix-org/synapse#12930 for more context.

[clokep]

According to @turt2live:

looks like web largely uses it in tests fwiw
the code which actually uses it treats it as optional

---

Although it treats it as optional, it is unclear to me that it will do the correct thing if it is missing.

[t3chguy]

This will regress element-hq/element-web#10200 or add an extra API call.

[clokep]

This will regress vector-im/element-web#10200

@richvdh @turt2live Any thoughts on this? Are the solutions either:

• Take the extra API call.
• Write a separate MSC?

[richvdh]

sounds like an extra API call wouldn't be that bad?

[turt2live]

An extra API call is the least of that dialog's worries. If we put more effort into edits, we can always write an MSC.

---

As an aside, is there a specific team we're expecting to action this? The js-sdk issue tracker tends to get, err, forgotten about.

[richvdh]

As an aside, is there a specific team we're expecting to action this? The js-sdk issue tracker tends to get, err, forgotten about.

It feels like the EW platform team would be best placed to deal with it? I know EW isn't the only user of the js-sdk, but EW is by far the biggest driver for the js-sdk's development.

(And, frankly, it'll be EW who get bitten hardest when Synapse stops grandfathering the unstable endpoint.)

[t3chguy]

The WAT is very unlikely to have bandwidth to do this in this quarter, the quarter plan is already somewhat assuming we get more bodies, given this is used by threads, maybe the team working on threads?

[richvdh]

@justjanne has volunteered to take a look at switching this over to the stable endpoint.