Ben Darnell

Ah wait, there's this: https://github.com/openssl/openssl/blob/6e94b5aecd619afd25e3dc25902952b1b3194edf/CHANGES#L236 Although that comment specifically refers to blocking IO and it looks like python has set this flag by default for a long time. Or there's...

> Could you point me to the code part where to look for finding the relevant commit to cherry-pick it? There were a lot of changes to iostream.py between 4.5...

> Do you have a patch for v4.5 series that limits the context to TLSv1.2 available somewhere? The workaround, as described in https://github.com/tornadoweb/tornado/issues/2536#issuecomment-447589432, is to set ctx.maximum_version = ssl.TLSVersion.TLSv1_2 wherever...

> How does Tornado implement things? Does it have non-blocking I/O implementation? Or is it blocking I/O? or both? Tornado always uses non-blocking I/O. And SSL_MODE_AUTO_RETRY is supposed to be...

The status is the same as it was in this comment: https://github.com/tornadoweb/tornado/issues/2536#issuecomment-490318306 We have a workaround (disabling TLS 1.3) but haven't identified the root cause or a proper fix. Upgrading...

Closing this as obsolete since it seems to only apply to Tornado 4.x.

`run_in_executor` returns `Awaitable` because that's what the corresponding `asyncio` method is typed to return. If I'm following the various threads correctly (starting with https://github.com/python/typeshed/issues/3999), now that https://bugs.python.org/issue40782 has been fixed...

Typeshed now has this method returning a Future, so this is a change we could make next time we upgrade mypy/typeshed.

This is related to #89. The rules for urlencoding are complicated and it's not correct to simply urldecode the whole path. For example, the urls `http://example.com/foo/bar` and http://example.com/foo%2Fbar` are two...

> says that at least the unreserved (ALPHA / DIGIT / "-" / "." / "_" / "~") characters should be decoded when comparing. I think this is not even...