[erlang-questions] relx extended deps

[e@REDACTED]

> One of the more interesting solutions that was discarded was to make
> ranch_ssl its own application.

it is actually the worst solution, for the reason of "unnecessary 
duplication" (the worst evil ever in IT) and for the reason you have 
just described below.

> But this only fixes Ranch and moves the
> problem one level up to Cowboy, which requires SSL for SPDY and soon for
> HTTP/2. These are the future protocols of the Web and Cowboy is aiming
> to support them at an equal level to HTTP/1.1 and Websocket, so it makes
> little sense to split HTTP/2 out.

> The solution retained is to fix the .app file, and document a way to
> remove SSL if you don't want it.

Seems to me rather a workaround than a solution.
It is acceptable, as well as it is acceptable to "sed" the .app file 
from within the Makefile, but a proper solution must not involve manual 
labour, especially when it is tinkering with a 3rd-party software (what 
is provided as a dep ideally must work without tinkering)

Also, i want to emphasize, that this problem has an *ALGORITHMIC* 
solution, which automatically moves the problem from the realm of humans 
to the real of computers -- no human should be forced to do algorithmic job.