Eitaro Fukamachi

Thank you for reporting. I'm still really not sure, but perhaps this should be done somehow in Woo socket code. I'll look into it and try to fix it.

Can you retry it with Woo's [allow-multithread-write](https://github.com/fukamachi/woo/tree/allow-multithread-write) branch?

Hm, that's strange and sounds like a bug.

There's no option for it right now. At a glance, `usocket:socket-connect` can take `:local-port` that looks what you need.

> If you get a chance, do a pull on cl-libuv and cl-async and give it another shot. I'm interested if your throughput goes up (and interested to see another...

Woo calls `as:write-socket-data` for each header lines. It might be a problem? I tried changing it to reduce the number of calls by creating a byte vector of the whole...

> The only thing I can think is...libevent is a lot faster than libuv. I never thought that but if so, the profiling result makes sense. > I wonder if...

Oh, I can't wait to try it! > Also, do you know when fast-http will have multipart parsing? Once that gets in, I'll happily switch Wookie over to fast-http/quri. Multipart...

Odd thing happened. ### With the latest Woo (rev:dae3d0) #### Before It seems after I changed Woo for reducing calls of `as:write-socket-data`, the `uv` branch's score was almost the same...

I changed Woo a little bit to call `as:write-socket-data` twice for headers and body (not once), then the score went back to almost the same to before's. Interesting. No, it...