> Thanks for the report! Not including matplotlib by default was > intentional, actually: If you install Flent and only want to run tests > (not plot them), you can do without and matplotlib does, as you rightly > note, pull in a lot of dependencies.
In my view, there are two classes of potential users for flent: Those running it with the intent to analyze the raw data, and those who simply want to see how their network(s) are doing. Without the plotting, you only have access to the raw numbers in an XML file. You would need to be using similar tools on a regular basis to make it worth your while to setup some means of interpreting the data. I'd imagine these users would be the hardcore base group of flent users. Most users who would try flent simply want to run a command and see how their system is running. Perhaps repeat the test next day/week to look for changes. The plotting capability gives them a visual way to see results without needing to decode the data format and perform custom analysis. Perhaps another option for you would be to offer two separate install packages - flent (which would pull in flent, netperf, and python-matplotlib) for the average user, and flent-core (which would pull in flent, and netperf only) BTW, the scale chosen/used for the plotted throughput is difficult to interpret for low speed connections. Scale shown on the plot for me is 10^0 to 10^-1 Mbits/sec with no dots or dashes to indicate where on the vertical scale those points happen. . Now granted I can squint and imagine subdivisions, but again, for the average user, showing 100 - 700kbps with all the dashes for 200, 300, 400, etc would be easier to interpret... Ping times show more subdivisions so it is easier to interpret (but still no dashes/markers). Actually, I just found the option in Preferences to disable log scale... I'll add this as another suggestion to have it disabled by default... Again, the wider audience doesn't really understand exponents...