Ping

[juan]

this happens when the server is running g_trueping 1.  as stated above, this isn't even really possible in a lan setting.  if you want a better estimate of the person's ping check splatterladder

[menatwork]

this happens when the server is running g_trueping 1.  as stated above, this isn't even really possible in a lan setting.  if you want a better estimate of the person's ping check splatterladder

 

Actually it is possible in a LAN setting; but it depends on the lan. If we're talking about a complicated LAN that's actually combined via OpenVPN and the likes - then yes, it's unlikely.

 

But on an actual, small-LAN with a *decent* router and properly configured firewalls (i.e. no rate-limiting, packet dropping on ET's port/UDP) and such - it's more than feasible. For instance - a ping (ping -c 100 10.0.0.3) from my laptop (cable connected) to my desktop (HTPC) downstairs yields this results (for clarity, we're talking about 100 pings sent/received):

--- 10.0.0.3 ping statistics ---
100 packets transmitted, 100 received, 0% packet loss, time 99000ms
rtt min/avg/max/mdev = 0.228/0.315/0.378/0.026 ms

So at best we've got a ping of 0.228 ms - on average, 0.315 and a max ping of 0.377ms - and a 'moving standard deviation' of 0.026ms (more on standard deviation here: https://en.wikipedia.org/wiki/Standard_deviation ).

 

Now adding the overhead of ET on Lan - one could surely get pings below 1 even unless I'm missing something (besides not taking into account how ET actually calculates it - but putting that aside, pings below 1 or slightly higher than 1 are most certainly possible in LAN situations - it also depends on the amount of clients; the computing power of the server; the networking equipment, etc).

[juan]

watch a demo from a european etpro match on lan, pings usually fluctuate between 2-12, and fluctuate more often than what you see on an internet server set at g_trueping 1.  even quakelive matches at quakecon display higher pings than what's stated by the servers, and i imagine they use the best hardware possible.

[menatwork]

I was actually talking about a small LAN party with a few friends (i.e. [player] -> router -> [server]); though from your reply I'm guessing you're referring to bigger events (quakecon, etc) in which case there's probably a little more to it than a single router. They're probably using at least a dozen or so switches and so forth.

 

Now on a larger LAN; i.e. Quakecon and the likes, I'd expect to see 2 or 3 hops at the minimum (i.e. for example: [player] -> [switch] -> [router] -> ([switch]) -> [server] | depending on where you're sitting and how they've configured things) and I would also expect a whole lot more bandwidth utilization (i.e. people downloading stuff, watching YouTube, playing online, other things) which will essentially (comparing to water going through a pipe) 'fill the pipes' more which definitely explains the 2 - 12ms pings.

 

In the end - like you said essentially, there's still the matter of how ET calculates it; I'm to lazy to actually look at the source code right now and it doesn't actually matter.

[SunLight]

In the end - like you said essentially, there's still the matter of how ET calculates it; I'm to lazy to actually look at the source code right now and it doesn't actually matter.

I'm too lazy as well (btw Silent source is not available, but it's based on etpub and they can't have changed everything - actually quite the opposite most probably).

 

But if we consider that:

- The server goes at 20fps so the timestamps obviously go in steps of 1000/20 = 50ms

- The client generates a command every frame, and they aren't necessarily sent immediately because of cl_maxpackets.

 

Then obviously there is some 'granularity' here, I don't know how to put it in English. It's not a case where you answer immediately. When you send a getstatus to the server, and you check how many ms pass before you get an answer, that is probably very close to the real ping (but then you can just ping the server, if you really wanna know the ping )

 

But not the 'ping' in the scoreboard, which should be actually higher than the real ping - and in fact it is in the standard ET. I don't know what that 'true ping' does, but it must still be calculated from the timestamps in the server frames and client commands, because no extra packets are sent aside from those ones.

[menatwork]

You're right - I'm guessing 'true ping' is actually a formula that calculates between the timestamps and then introduces some minus-magic to make an educated guess to the actual ping. What I do know (if memory serves me well) is that true ping actually updates the value - continuously, as in - you'll see the players ping actually fluctuate a lot from a ping of 2 to 20 to 30, etc.. it's in constant 'flux' whereas without true ping, it's actually just an average (iirc!)

 

It's a shame though that Silent's source code isn't available especially since they based it of pub and that source is still available. I wonder what the thoughts were behind that decision - it feels silly though, they took pub's source modified it and then essentially made it closed source (unless of course it's source is somewhere public - where we can't find it, but does that really change anything?). I wonder how they draw new contributors - it'd be awesome if it was up on GitHub.