Mikhail Korobov

### Resource deallocation This is kind of important because even in this code: ``` python resp = yield Request(url1) resp = yield Request(url2) ``` the first `resp` value lives in...

Obvious observation: an alternative to ``` py resp = yield Request(url1) del resp resp = yield Request(url2) ``` is ``` py resp = yield Request(url1) del resp; resp = yield...

Yeah, it is not specifically about inline_requests. I think we shouldn't _change_ Scrapy API from callbacks to coroutines, we should provide _additional_ coroutine-based API: while looking nice couroutines have their...

Hey @patiences! As async generators are implemented in Python 3.6, I think that's fine to use them instead. With async generators this code would look like this: ```py class MySpider(scrapy.Spider):...

FTR, `await scrapy.Request(url1)` syntax is hard, because in this case scrapy.Request needs to schedule itself - it needs a reference to Crawler, which it currently doesn't have by design. As...

> Do you lean towards either option? I might prefer to have await rather than async with just because it feels more idiomatic to see them together. @patiences the difference...

> How are we looking to maintain backwards compatibility ? Are we going to generate this as a separate API, from scrapy ? And then figure out how to backport...

Nice summary, thanks @wRAR!

Some thoughts on the questions: > Should an exception be raised or returned in some form? Looks like raising is better for the API user. +1 to raise an exception....

A good point about prioritizing in-memory request @Gallaecio, I think you're right. I don't see other caveats here. Maybe we'd need to make sure that such requests are not seen...