[ENet-discuss] roundTripTimeoutLimit with big ENET_PEER_TIMEOUT_LIMIT

Nuno Silva little.coding.fox at gmail.com
Fri Nov 6 09:48:51 PST 2009


I just remembered, you could just get the server to tell the client that a
peer disconnected. That way, it'd skip it's attempt to connect to the
client.

On Fri, Nov 6, 2009 at 5:03 PM, Uj <ujdevil at gmail.com> wrote:

> Well it's your application do decide if the disconnection is sudden...
> (expect for crashes of course, but your app shouldn't crash ^^)
> I can't see Alain the difference in your case between disconnection and
> timeout actually... timeout causes disconnection.
>
> If those settings are for debugging purposes, then take them off in the
> release version and it should be fine :p
>
> If you can spot the moments when you need high timeout,
> then set them dynamically and adjust them when needed.
> Or use specific messages to prevent specific situations (like asserts or so
> on).
>
> Hope this helps,
> regards,
> Ju
>
>
>
> On Thu, Nov 5, 2009 at 7:51 PM, Alain Becam <Al at thegiantball.com> wrote:
>
>> Indeed...
>>
>> 2009/11/5 Nuno Silva <little.coding.fox at gmail.com>
>>
>> Not to mention you'd have the same issue if the users' network was
>>> disconnected suddently.
>>>
>>> On Thu, Nov 5, 2009 at 5:27 PM, Alain Becam <Al at thegiantball.com> wrote:
>>>
>>>> Hi,
>>>>
>>>>    Thank you for your answer! Actually when they disconnect properly, there is no problem. It is in case of crash that we have this problem. Unfortunately it happens ;)
>>>>
>>>> //Alain
>>>>
>>>> Hi Alain,
>>>>
>>>> Is it possible for your peers to send a disconnection message before going
>>>> away ?
>>>>
>>>> regards,
>>>> Ju
>>>>
>>>>
>>>> On Thu, Nov 5, 2009 at 2:03 PM, Alain Becam <Al at thegiantball.com <http://lists.cubik.org/mailman/listinfo/enet-discuss>> wrote:
>>>>
>>>> >* Hello,*
>>>> >*      Greetings from a new happy user. We are using ENet has the "low-level"*>* communication API for an object-oriented solution, with object replication.*>* Thanks for this great library !*>**>* Mostly for debug purposes, we increased the TIME_OUT, like that:*>**>*    ENET_PEER_TIMEOUT_LIMIT                = 3200, // 32*>*    ENET_PEER_TIMEOUT_MINIMUM              = 5000, // 5000*>*    ENET_PEER_TIMEOUT_MAXIMUM              = 300000, // 30000*>**>* It works well to avoid too quick disconnections, but we still have a*>* problem. When a new peer connect, it receives the list of existing peers*>* from the server, then try to connect to all peers, and when done, send an*>* acknowledgment to the server. Then the server replicates the existing*>* objects. But with these time-out, a disconnected peer is discovered after a*>* very long time, and if a new peer connect, it will never send the ack to the*>* server (because he is waiting for the connection and until it receives the*>* disconnection event).*>**>* We tried to do that in protocol.c (line 1270):*>**>* outgoingCommand -> roundTripTimeoutLimit = (ENET_PEER_TIMEOUT_LIMIT/100) **>* outgoingCommand -> roundTripTimeout;*>**>* But it seems to have some nasty side-effects (when some communications are*>* occurring, it might disconnect very quickly again)...*>**>* Any way to have a quick disconnection and still a long time-out for comms ?*>**>* Kind regards,*>*     Alain*>**
>>>> >* _______________________________________________*>* ENet-discuss mailing list*>* ENet-discuss at cubik.org <http://lists.cubik.org/mailman/listinfo/enet-discuss>*>* http://lists.cubik.org/mailman/listinfo/enet-discuss*>**>**
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> ENet-discuss mailing list
>>>> ENet-discuss at cubik.org
>>>>
>>>> http://lists.cubik.org/mailman/listinfo/enet-discuss
>>>>
>>>>
>>>
>>> _______________________________________________
>>> ENet-discuss mailing list
>>> ENet-discuss at cubik.org
>>> http://lists.cubik.org/mailman/listinfo/enet-discuss
>>>
>>>
>>
>> _______________________________________________
>> ENet-discuss mailing list
>> ENet-discuss at cubik.org
>> http://lists.cubik.org/mailman/listinfo/enet-discuss
>>
>>
>
> _______________________________________________
> ENet-discuss mailing list
> ENet-discuss at cubik.org
> http://lists.cubik.org/mailman/listinfo/enet-discuss
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cubik.org/pipermail/enet-discuss/attachments/20091106/518b2919/attachment.htm>


More information about the ENet-discuss mailing list