gping: Ping, but with a graph

I found that tool from Ben Kuhn's blog, and at first it seemed great. But I've seen some weird behavior from it. Examples include drastically different behavior on a refresh (and not just for the 'blue' website, also for gstatic.com) and pings to Madagascar which are apparently faster than the speed of light. So now I'm somewhere between confused and distrustful.

If anyone has a detailed guide on using gfblip I'd be really curious to read it.

That's a fun link.

Intriguing, but: I don't think it's working as intended from my Brave browser.

After about 2 seconds of blue/green pings that seem faster than should be possible (<10ms), it goes full red.

It does work in Chrome! (Perhaps, Brave is doing some faster-fail on the probe request from the beginning, and then a cached-insta-failure after ~2 seconds?)

FYI - if you're getting a max red line on this site make sure you're not using the https version (and you may need to whitelist it in https-everywhere).

Regular old time series are such a powerful tool!

My green pings to gstatic are steady, 19-21ms.

My blue pings to calgary are reliably ~2000ms then ~200ms. I wonder where in the chain that pattern is originating. I wonder if it'll stay like that over time.

Comparing performance of this page between Chromium and Firefox is just very depressing :(

But therefore you need perl, gnuplot and feedgnuplot, the latter isn't even available via package manager on my distro.

Fantastic, thanks, I think this was the final push I needed to start abusing gnuplot.

Though for some reason the plot only updates when I give or take the plot window focus (refreshperiod mentioned in the man page doesn't seem to help, and I think it should default to 1Hz).

Another great example how a couple of traditional unix tools (if you want to count perl and gnuplot as traditional unix tools) can replace 250 lines of high-level language code (https://github.com/orf/gping/blob/master/src/main.rs).

I like how the more recent versions of mtr have a coloured heat map, showing the 3D data, essentially.

(x=time of test, y=hops, z/color=ping time)

If you pass the -z switch to mtr, it also shows the associated AS.

The Cargo metadata says MIT.

Well pings will have a bit of a weird distribution. If you did a histogram of your ping for a bit you would see most values sitting near some "floor" of X ms and then a long right-sided tail of values much larger than X ms. It seems to look sorta like a gamma distribution from a quick test of my wi-fi ping. The mean would sit somewhere to the right of the largest mass of ping times. If you wanted to know where most pings were you could look at the median (which will sit right where most pings are). If you were interested in the really large values of X ms then you could compare the mean and median to find out just how much those slow ping responses are dragging your mean away from the median.

I didn’t put too much thought into the name. It’s ping, with a graph. Graph ping. gping.

nping does sound kind of interesting, but meh.

This is unlikely to change since if ping changes many programs would broke. It is kinda like ipconfig.

AFAIK, ping is a setuid binary (since it needs raw socket, that is a privileged operation) so a program that implements ping by itself would need to do the same (or it would work for root).

I’ve just updated pinger to use a windows-specific API that doesn’t need root. But yes, in general it’s a nightmare especially with ping executables that don’t report timeouts at all.

This is a great example of one of the failings of the UNIX model. You can’t really parse the ping output across distributions easily, it’s built for humans to consume and until recently didn’t even report timeouts to stdout. Not to mention needing to invoke “ping6” for ipv6, which I need to add support for.

However it’s the best way to do this without needing to run gping and root.

A great project Brendan. But as stated in Github issues, it is not worked on newer version of Go. I'm trying in Windows 10 and get error which look like:

`Error: 'date-time' main.go:59: open C:\Program Files Direct\duck\pings.json: The system cannot find the file specified.`

`not supported by windows. Request timeout for icmp_seq xx`

That’s a really great idea. Not hard I don’t think

This relies on libraries too.

I reckon a D implementation would be a little bit shorter, I might try and write one tomorrow.

mtr makes for a pretty decent gtraceroute