I am using http_transfer to communicate with a server running locally which does some work, then responds with some data. A typical round trip is < 0.1 seconds.

I find that when the data argument to http_transfer is larger than about 1000 bytes, every call blocks for 1-2 seconds before my http server receives the POST.

This issue also affects http_post.

I suppose that the reason might be the MTU of 1000 bytes. Above that limit, multiple packets are sent.

Good gues but the delay doesn't scale linearly.

At the moment it looks like some bad interaction between Zorro and Python SimpleXMLRPCServer. I'll do some more experiments and let you know if I find anything. But my visibility is limited.

As to my knowledge, Zorro uses the CURL library for http posts (http://curl.haxx.se/)