Reynir Björnsson

We use `Dream.path` in two cases: - `.../build/latest/**` where it redirects to `.../build/:build/**` preserving the tail of the path - `file/**` which looks up in a database to find the...

I'm not sure the options string is important anymore. The `--opt-verify` configuration directive is marked as deprecated. For completeness and to avoid annoying warnings it could be nice to implement,...

I tried to figure out why we decided to disallow that. I have not found good arguments why. There are some issues that affect server side which are then mitigated...

So I took a closer look, and the wire packet format for tls without either tls-auth, tls-crypt or tls-crypt-v2 differs by not having what openvpn-rfc calls `replay_packet_id` (in our code...

Brave man ;-) This part of the openvpn-rfc is relevant: https://openvpn.github.io/openvpn-rfc/openvpn-wire-protocol.html#section-5 This is the code dump of how far I got. It might be a good starting point (or maybe...

Thanks! This unblocked me! I'll propose a PR soon.

I couldn't find one place where it's nicely documented, but you can already tell from this snippet: https://github.com/mirage/mirage-solo5/blob/be95f5a5b1af6d19ce60c1ab2475d27c007f824a/lib/main.ml#L58-L67 The ready set is a 64 bit integer used as a bit...

I was not feeling so confident about the number 63. We have `MFT_MAX_ENTRIES` defined to 64: https://github.com/Solo5/solo5/blob/ff63477507a032583e2e2a91dc93eed5b27103cd/include/mft_abi.h#L87-L90 And in `mft_validate()` we require that the number of `entries

I wrote a benchmark which suggests there is a small improvement, but it is probably too noisy to draw conclusions from. Source available here: https://gist.github.com/reynir/b8ff35fa7de0dc880522ff83a2bde04d Edit: I found I was...

I realized I was benchmarking the wrong thing: with that setup I/O is probably always available and the `wait_for_work_on_handle` is not exercised. So I rewrote it to exercise that function....