[ENet-discuss] Safe to increase ENET_PEER_PING_INTERVAL?

[John Wood]

Ok that's interesting, thanks for the info!

So are you saying that I should NOT increase it to 5000ms?

John

On 21 May 2010, at 19:07, Lee Salzman <lsalzman1 at cox.net> wrote:

> The pinging I mostly put in to monitor the connection health from  
> round trip times. Normally, reliable packets are getting a round- 
> trip acknowledgement, which gets used to calculate the throttle. But  
> if those aren't sent for a while, then the ping is sent instead so  
> that the throttle can still keep up to date, and also to see if the  
> connection is still alive. If it helps with firewalls and other  
> things, then that's an unintended benefit.
>
> Lee
>
> John Wood wrote:
>> Hi there,
>>
>> I wanted to know if it's safe to increase ENET_PEER_PING_INTERVAL,  
>> which determines how often a ping is sent
>>
>> Background: I'm using e-net for reliable data transfer only (no  
>> unreliable packets).
>>
>> I noticed it's quite "chatty" - sending a ping every 100ms  
>> (ENET_PEER_PING_INTERVAL).
>>
>> I wanted to reduce this chattiness, as I'm not sure I need it. I'm  
>> guessing that it's chatty a) to act as a keepalive to make routers  
>> keep forwarding packers and/or b) to update info about metrics e.g.  
>> RTT.
>>
>> So I increased ENET_PEER_PING_INTERVAL to 5000 and it works  
>> perfectly. I even tried 20000 and it still works (though I think  
>> this is too high as we run the danger of aggresive firewall/routers  
>> thinking the "connection" is dead).
>>
>> My question is this: is there any other unintended consequences of  
>> increasing ENET_PEER_PING_INTERVAL to, say, 5000?
>>
>> It's only used in the code once, for determining when to ping, but  
>> I wondered if other things might rely on that hard coded value.  
>> E.g., will peers "time-out" too regulalry now? Will it break some  
>> of the "latency" metrics e-net gathers?
>>
>> I can't see any problems, but I want to be sure.
>>
>> Regards,
>>
>> John Wood
>
> _______________________________________________
> ENet-discuss mailing list
> ENet-discuss at cubik.org
> http://lists.cubik.org/mailman/listinfo/enet-discuss