Bryan Montz

Yes, but the problem was that it was creating and throwing away a socket every time a RelayConnection was allocated. I thought of a better way that eliminates the optionality...

Right. Super critical code. I'd love for it to be as clean as possible and have tests.

I agree. Maybe we could even have it so that we parse the https://damus.io/note_identifier link and render it as @note_identifier instead so that it doesn't do the snort social preview...

@alltheseas I'm working on this. 🙂

Fixed in this PR, but initially Will wanted to fix it a different way. I wonder if he'd reconsider, at least for now. https://github.com/damus-io/damus/pull/896

This is excellent! 👍

No, I'm not sure how they're triggering Apple's translate sheet, if that's what's happening. I googled around for a minute and didn't see a way to do it.

If those delays are too short at the beginning, the exponential backoff algorithm's constants can be tweaked and tuned to your liking. One example might be that the first delay...

Since you said it needed work, I took the liberty and added some refinements to `RelayConnection` and `RelayPool`. Some highlights: 1. All event decoding is done off the main thread....

Next I plan to add some tests for reassurance.