Andrew Svetlov

I don't think that the library should manage the Redis server limitations. It can unpredictable side effects in general, better to avoid such things and keep the library logic close...

I didn't write that this is a configuration logic but "Redis server limitation". Should aioredis be a thin wrapper around the Redis protocol or a fat library with many potential...

@gvanrossum I don't see any harm from #305 aiohttp (and all other libraries supported by me) definitely doesn't change own behavior after PR appliance.

Definitely better to fix Popen by, say, adding a flag for disabling read. Honestly I doubt if anybody pass `preexec_fn` to asyncio subprocess calls. Under high IO pressure the whole...

Ok, but most likely it affects Popen also. Overriding the whole `_execute_child` scares me.

I think the PR makes sense: we raise `ResourceWarning` for unclosed transports in debug mode. I believe we should do the same for unclosed loop too.

In aiohttp we have `limit` parameter for `.readuntil()` and family. @gvanrossum is right -- we need custom streaming for processing compression and chunked encoding. I'm not 100% happy with current...

Agree, per stream limit is enough.

+1 for `spawn`. Perfect name.

I believe @gvanrossum wrote about returning a future from `.close()`. But diagnostics for mandatory calling `.wait_closed()` after `.close()` is even worse, isn't it? Right now I'm making every `.close()` function...