Opening a connection may be the most OS/deployment portable thing you can do to
check for Wide Area Network access, but it is probably harder to optimize for
network latency and the timeout for establishing connections can be pretty long
- seconds even.
Here is a faster idea that might only require a little work to make
portable-ish. With `ping` you can crank the time-to-live (TTL) down and endure
just a few hops to test (on Unix `ping -c 1 -t 2 1.1.1.1` on Windows `ping -n 1
1.1.1.1 -i 2` {untested}). For me, I get down near 1.0 milliseconds since we
limit the number of hops to "close by", not "all the way to 1.1.1.1".
You can run external commands in Nim with `os.execShellCmd` and (on Unix)
detect via the exit status if you even made it 2 hops (or however many you
like). If you did not, you have no WAN access. You could even exit with a nice
message about checking your local net setup. (You'll have to check what Windows
ping does Re exit status codes.) If there is no ping or you cannot run it, you
could always fail as if there is no network.
* * *
You could probably write some Nim code with raw sockets to do an API version of
the above idea. The problem here is that this usually needs certain
administrative privilege/setuid-programs that may vary in portability across
OSes even more than CL options (or the name of the command).
You might be able to set socket options to crank down the timeout on your
connect. This may improve things some, but will still always do "all the hops"
to the destination" (at least 1 full round trip). So, it will always be
order(s) of magnitude slower than my ICMP ping packet suggestion.
