Erik Johnston

Seems like a bunch of people are vaguely happy with this, so... @mscbot fcp merge

@mscbot fcp merge

What's the status of this? There is currently code in Synapse to make this work that is on the hot path, and we should rip it out if we're not...

Is this actually being driven by something? If it's not a feature that people are asking for then I'm not sure there's much point finishing it?

I think this was disabled because it had a tendency to take out synapse, especially when the freenode bridge randomly did it. It may be less of an issue if...

> honestly if the freenode bridge hits `/sync`, we've done something terribly wrong. Can we stop optimizing the proposal for a bridge that won't use sync? There's nothing in this...

Operational issues are caused by the fact the spec says we must support unbounded data fetching requests, and so the spec can fix those operational issues by not mandating we...

> One use case would be go-neb: it doesn't use appservice transactions but fits really nicely for reserving a namespace of users. Because the spec says that the `sender_localpart` is...

> Bridges don't need to be big. This is meant as a first step for user-hostable appservices Absolutely true, but they do often have a tendency to grow big after...

@mscbot concern whether this is something we should be encouraging ─ https://github.com/matrix-org/matrix-doc/pull/2190#discussion_r595444852