Claire Foster

> What I was essentially suggesting is that you reinterpret to 7 x P matrix of Float64 to get a P-length `Vector{SVector{7,Foat64,7}}` and that `getindex(::Vector{

Yes we should try those out when we get the chance. More immediately (julia 1.4), we can try the idea of copying into an appropriately-shaped container of `VecElement` and seeing...

Great summary. I've hit this a couple of times in the opposite order: 1.0.0 -> 1.0.1 -> 2.0.0 where 1.0.1 is a tiny release made on a branch, but gets...

A version of this could be made to work and would be very cool. What would the API be? How about having `y = @remote x` "just work" when `x`...

> What if the user really just wants to take the Observable x from the remote and not create a new y that tracks it ? I guess they can't....

Very interesting, thanks so much for the thoughts. It's great fun to finally have a proper design discussion about this and related issues (here, in LoggingExtras and in Base). >...

> You can always convert sink/source-oriented APIs back and forth as long as there is "adjoint" relationship (this realization led me to this PR [JuliaLang/julia#33526](https://github.com/JuliaLang/julia/pull/33526) that suggests to convert iterator...

> Because of the argument order, we need to manually overload by defining the other two methods No problems. I like the current argument ordering anyway by analogy to other...

I've fixed the conflicts by rebasing and pushing to the pull request branch. That also seems to have triggered CI. Can you have a look at what's gone wrong there?

Oof. I suspect I've seen that finalizer error before. Probably not your fault! It's hard to clean up the ssh tunnel cleanly when the process exists. (Unless there's a way...