first question (metric for protocol): That would defeat the purpose of SAGA.  We would also then 'need' a metric which tells you the file system type 
a file lives in, the queing system used by a job service etc etc.  Is there a use case for this?

second question: yes, proxies seem to be outside the scope of SAGA right now.  Should that be changed?  explicitely documented?  That holds for all 
communication, not only streams!  I'd argue that connection setup is a backend problem, and should be solved/configured on system level, i.e. out of 
bound of SAGA.