NTPsec's longer-term objectives

[Eric S. Raymond]

Hal Murray <hmurray at megapathdsl.net>:
> It's worth considering writing something like a ntp-lite in Python.  It would 
> probably take a handful of shims to get at OS calls that Python doesn't 
> support directly.
> 
> Consider a client-only, no refclocks implementation.  The timing  parts are't 
> that tricky.  The OS provides a receive time stamp.  On transmit, we could 
> grab the time before and after the interesting chunk of code and add the 
> difference to the error budget or try again if it is too big.

While this is in some ways a tempting thought, I think the energy we
might spend on this would be better directed towards a newer language
that *is* suitable for soft realtime and has good provability properties
for security.  Rust or Go seem like the obvious candidates.
-- 
		<a href="http://www.catb.org/~esr/">Eric S. Raymond</a>