I think we need two config options - but they set the same variable internally and only call ONE curl interface - the one you add here. We need also to verify the http_client API if we need to add something there. Having two variables internally is just overhead.
Alright, I'll try something like you propose.
I had thought of this initially, but it seemed too convoluted because the configuration can come from 3 different sources : global parameter, httpcon parameter, or file configuration... so each time we need to "fixup" the actual timeout, depending on whether we have the ms timeout or not.
I'll try a new pull request, and then you can check if you think it is better. Thanks for the feedback!