Empty path shall result with /, not with an empty string. Even if it is
handled by some backends, it's a kind of a lie, especially, when additional
path from the request come in.

XXX - is that really the place to fix it?

XXX:
 * improve it more
 * adapt tests

Test backend needs to mimic all possible real backend cases.

Just an example of how tests shall be rewritten.

Also backend is backward compatible in old bad way, with representing
headers as a dict instead of sorted list of lists.

Fixes handling failed backups and default_exception_handler.

See merge request nexedi/slapos!1685

Stop using deprecated ssl.wrap_socket, and let unset key-file mean
the key is included in cert-file, as is done in python ssl module.
Also let the server crash if specified cert-file or key-file does
not exist, instead of silently falling back to non-TLS http.

Add 'socketpath' and 'abstract' options for path based unix sockets
and abstract unix sockets respectively.

See merge request nexedi/slapos!1716

See nexedi/neoppod@f4a5fe3b.

This was sometimes causing false positives, for example if site have
only one hit and this hit takes more than one second, apdex is 0% and
promise failed, it seems useful to at least have a way to disable.

We are discussed on nexedi/slapos!1716 (comment 225941)
an idea of introducing a "minimal number of hits" parameter to not
detect this false positive on sites that are not accessed enough to
calculate a meaningful apdex.

Needed for password authentication to websocket
A bug was discovered in previous version

See merge request !1717

used in theia or with security vulnerabilities (idna & Werkzeug)

With fixes for two critical CVE