Whenever an IPP client is configured to use a proxy, it would probably
make sense to have the client send
"POST http://foo.bar:XXX/zot HTTP/1.1" to the proxy when attempting
to talk to the ipp object "ipp://foo.bar/zot".
As far as I can tell from a very casual analysis, this is the only
place where it would be necessary to actually send the string "http:"
to refer to a ipp object. Every other URI that refers to a ipp object
could use "ipp:" instead.
I don't see a problem with doing things this way, as long as it's
clearly documented. Perhaps it would be wise to add a section called
something like "Tunneling of IPP requests over HTTP proxies" to the
protocol document that specified such details.