move sewer to asyncio

[komuw]

most of the things that sewer does are network IO related(making api calls to letsencrypt, making other calls to DNS providers etc)

This is precisely the kind of work that python3's asyncio really shines at.

I've had some experience now working with asyncio and it is awesome;
naz: https://github.com/komuw/naz :: an async SMPP client
wiji: https://github.com/komuw/wiji :: an upcoming async distributed task processor.

probably also release a new major version because if we are going to so asyncio, we better do it with python3.7+ and that may be a breaking change for some.

And if we are doing a major version with breaking changes and some async goodness, we might as well also do: #91 DNS changes )

[mmaney]

I'm not sure what asyncio would buy sewer, unless you were processing a bunch of separate requests in parallel (which the cli certainly doesn't support). Given how much the API forces sequential access (must get nonce in order to post this, another nonce in order to poll for completion of that, etc.), using request's session to make best use of persistent connections might be of more value. It would at least get rid of the User-Agent repeated for every request.

Caveat: I haven't used either asyncio or requests much before I fell into working on getting rfc8555 working here. :-)

[komuw]

yeah, there maybe are more lower hanging fruits.
asyncio would make a lot more sense if someone wanted to batch certificate requests for different domain names.

[mmaney]

Heh, I was going to say something about batching being the one likely use case for this. Also that making sewer more parallel might push rate limits [harder].