Maximilian Hils

Ì think the problem here is unrelated to `.. include::`, it's much rather that we don't support `:ref:`. ``` :py:class:`Task` ``` works as expected for me. Is `:ref:` something that...

Yeah, I think that's extended rST syntax we don't support (yet). ``` :py:class:`Task` ``` works if Task is defined in the same file, otherwise ``` :py:class:`pydatatask.task.Task` ``` does the trick....

(Happy to look at this once you mark it as ready)

@thvdveld: Thank you so much! 🍰 I was just looking into this issue when upgrading smoltcp in https://github.com/mitmproxy/mitmproxy_rs, what a fantastic timing. 🤩 Thank you for your work! ❤️

This should probably be expedited in light of https://sansec.io/research/polyfill-supply-chain-attack. https://www.mathjax.org/#gettingstarted still references polyfill.io.

If we can only repro this with TLS 1.2, I suspect this is likely triggered by renegotation. A WireShark capture + SSLKEYLOGFILE may confirm that.

> @mhils : I believe this PR actually don't keep unneeded (encrypted) raw traffic in memory. I am pretty sure it does. But I guess this is acceptable for a...

Thanks for the clear issue description. The root cause is probably here: https://github.com/mitmproxy/mitmproxy/blob/1e5464a9e632b67df3d49d4e2e39e6b60eea5b0e/mitmproxy/proxy/protocol/http.py#L418-L449 We set timestamp_end before triggering the `response` hooks (`self.channel.ask("response", f)`), and then only update it later on....

We don't support TLS-PSK at the moment, although that is definitely something that is in scope for mitmproxy. We use pyOpenSSL/cyptography under the hood, so implementing this is a matter...

@not-duckie: there's https://github.com/mitmproxy/mitmproxy/blob/main/CONTRIBUTING.md and then you want to look at https://github.com/mitmproxy/mitmproxy/blob/main/mitmproxy/addons/tlsconfig.py I think. :)