Release: common@3

[josephjclark]

Release branch for common major 3.0

• remove map
• add new map
• tidy and update expandreferences
• remove axios
• export lodash

[josephjclark]

hmm. Since the util refactor is looking too big and complicated, maybe we should release this as-is 🤔

I'll try and look at the docs thing on #1173. If there's a solution for the sake of a couple of hour's work, we'll do it. Otherwise we'll close that issue and never do it. It may just be too expensive

[josephjclark]

I think we made a mistake with rapidpro here, and maybe others. But I don't want to take time now to fix them - we can roll in fixes later if it becomes relevant.

Basically rapidpro no longer exports http. That's wrong, it should continue to export http. But it's a different version of http because it's now exporting the common 3 http namespace, not the common 1 namespace.

Let's leave this for now, I just wanted to leave a note for the record

[josephjclark]

Also, on this branch, can we review all http exports?

Common exports some very low level, generic http helpers, They don't do any auth stuff for you and they must take absolute URLs. They don't use config. This is fine.

If an adaptor declares its own http helpers, those helpers are not generic. They're bound to that service. And so that adaptor must not export the common http namespace. Because it's confusing and the two conflict.

So now I'm in doubt whether any adaptors should export common http at all. It sends the wrong message doesn't it? We usually say that steps should not send external HTTP requests. But where we export common http we sort of allow it.

I want to think about this and come back to it later.

[josephjclark]

Going to veto the util namespace on this branch. Let's just focus on getting common 3.0 out of door.