We need to decide if trailing slashes are optional on requests

[turt2live]

See https://github.com/matrix-org/matrix-doc/issues/1944
See matrix-org/matrix-spec-proposals#2061
See matrix-org/matrix-spec-proposals#2097 (comment)

Personally, I'm of the opinion that it should be optional (ie: can be included or excluded).

[ara4n]

my thoughts on this are as per matrix-org/matrix-spec-proposals#1849 (comment)

[richvdh]

my thoughts are, fairly strongly, that there should be exactly one way to do it. Having two ways to do anything leads us to the situation where:

• most clients do it one way, some do it the other
• a new server implementation is written and tested against the common clients, but never against the less common clients which do it the other way.
• some clients are compatible with some some servers some of the time. oh hai xmpp.

[KitsuneRal]

Being a happy developer of a client that generates paths directly from Swagger spec (and we have at least one case of doing the same for servers) I obviously would like that the Swagger definition is strictly followed. That makes the case in matrix-org/matrix-spec-proposals#2061 somewhat stand out: Swagger mandates required: true for all path parameters so the only reasonable way to incorporate this change is, exactly as the proposal section in it states, to make a separate endpoint (another, unreasonable, way would be to say that keyId is required but can be empty which is very confusing).

I agree with Rich that for the sake of consistency across the ecosystem we have to follow exactly one way, and no trailing slashes seems to be a de-facto accepted convention requiring the least number of adjustments. I would also prefer to declare endpoints with trailing slashes as generally a no-go zone within Matrix API, to simplify tracking down bugs in clients where the last parameter on the path is empty.