So this is obsolete and not correct?
I've quickly skimmed through the document and I couldn't find the part which contradicts what I wrote in my previous post.
Then I went forward to check. My router runs ROS v7.8 (so the quick test results might be different when testing against other ROS version, e.g. v6). I enabled proxy on ROS, leaving everything else at default settings. Then I ran the following linux CLI command:
https_proxy=<router_IP>:8080 wget https://<insert your favourite FQDN here>
and contents was downloaded just fine. Then I re-ran the command but omitting the https_proxy variable setting and the contents of retreived html file was identical between the two runs.
The command above is first setting environment variable https_proxy to combination IP:port of proxy running on my ROS. And passing environment variables is the standard way of making system-wide proxy config for CLI commands in linux. There's another variable, named
http_proxy, which is used by CLI commands when URL requires non-SSL connection.
Both are similar to setting HTTP proxy and HTTPS proxy in various GUI web browsers (e.g. in firefox: Edit -> Settings -> General -> Network Settings -> Settings ... there are 3 different settings in manual mode: HTTP proxy, HTTPS proxy and Socks Host).
But, as I already wrote: using proxy for https connections only works where web client knows that it's using proxy server and uses HTTP to connect to proxy, then it asks proxy for web content ... And obviously in this case SSL/TLS is not end2end between web server and browser, its only between web server and proxy.
Some web proxy servers (e.g.
squid) support SSL for connections between clients and themselves. Not many web clients support this feature natively though. If this is used, then SSL connection is made in two legs: one leg is between browser and proxy, another leg is between proxy and server. This functionality is (AFAIK) not supported by ROS proxy service.