Investigate inaccuracy in speed reports about speedtest

[zpeters]

Rather than chasing this down any further, im going to focus on the new (undocumented) binary protocol. I believe this will both simplify things and lead to more accurate results. There is too much "magic" in their current way of calculating the upload/download speed and it's probably a better payoff to focus on the new protocol for now.

Will reopen and reinvestigate if the binary protocol turns into a dead end.

[zpeters]

I just put out a debug release that will add some extra info to the upload process. Would you mind running this with the "-d --uo" switches (-d = debugging, --uo = upload only) a few times and we'll see if we can catch this in the act again.

http://media.thehelpfulhacker.net/index.php?dir=speedtest/v0.8.4-d-1-ga02d067/

Thank you!

[ross-w]

FWIW we have the opposite issue - we are getting speeds considerably higher than our internet connection is capable of.

For example, on a 100/100 fibre connection:

$ speedtest-mac-amd64-v0.8.4-d
github.com/zpeters/speedtest -- unofficial cli for speedtest.net
Server: 2169 - Melbourne (Internode)
2016/02/24 09:19:41 Testing download speed
..........
2016/02/24 09:20:00 Testing upload speed
.....
Ping (Lowest): 2.30 ms | Download (Max): 330.05 Mbps | Upload (Max): 522.93 Mbps

The web version reports 93 down, 89 up.

Happy to run with debug switches but it includes information I'd rather not share in a public forum. Let me know how I can supply this publicly or if I can redact some of it without making it useless.

[zpeters]

Not a problem. If you want to email them to me privately it's [email protected].

[zpeters]

Still trying to find the time to look at this