[ENet-discuss] ordered packets

[Tibor Klajnscek]

Why don't you guys just use ENET_PACKET_FLAG_UNSEQUENCED for unordered 
packet delivery
without using so many channels? Just wondering if i'm doing something 
wrong :)

Luis Cabellos wrote:

>Well, in the server I don't discriminate among the channel that the 
>packet came from. So in the server I treat all packets the same way. 
>Enet use the channels to mantain the order of packets. All the packets 
>in the same channel are in-order, and if one of the packets arrive at 
>server too later, it will be dropping and wait for acknowledgment.So if 
>you know that the packets don't need the order, the use of a range of 
>channels
>
>Let me show one ASCII example :)
>
>    When you use on channel
>    Send:
>     1 2 3 4 5 6 7 8 9 10 11 12  -> channel 1
>
>    Receive:
>     1 2 [>> 4 5 3 <<] 6 7 8 9 10 11 12  ->channel 1
>     -> Enet drop 4 and wait for arrive of 3
>
>    And when you use 3 channels, [ min = 1 , max = 3 ]
>    Send:
>     1     4     7     10        -> channel 1
>       2     5     8      11     -> channel 2
>         3     6     9       12  -> channel 3
>
>    Receive
>     1   4       7     10        -> channel 1
>       2   5       8      11     -> channel 2
>             3 6     9       12  -> channel 3
>     -> And you get 1 2 4 5 3 6 7 8 9 10 11 12 without drop / wait
>
>So the client is the only that change the lenght of the range of 
>channels, with a long range less probability of out-of-order. And the 
>change of the number of channels if you maintain that ( min <= max ) it 
>will be always one channel at least. In the other hand, if you increase 
>the range, I think that the server will not slow down too much with the 
>use of more channels.
>
>- Luis | Tragnarion
>
>Paul Collier wrote:
>
>  
>
>>I'm curious about setting the range of channels. Is there any way to
>>set it other than by the client at connection establishment? If
>>someone tried to clone a program's protocol with enet but changed the
>>number of channels, could this possibly affect the behaviour of the
>>server you were running? (i.e. they set it to zero--could this break
>>the server network code? Or perhaps they set it to a very high
>>number--could this slow down the server?) Enet seems to be pretty good
>>about security from what I've seen, in any case.
>>
>>- Paul
>>
>>On 3/30/06, Luis Cabellos <lists at tragnarion.com> wrote:
>> 
>>
>>    
>>
>>>You can send Unordered packets using a range of channels and send the
>>>packets sequentially in those channels. e.g: If you reserve [min, max] for
>>>unordered send, the next code snip will work like an unordered send.
>>>
>>>    packet = enet_packet_create( msg, size, ENET_PACKET_FLAG_RELIABLE );
>>>    enet_peer_send( server, unorderedChannel, packet);
>>>    ++this->unorderedChannel;
>>>    if( unorderedChannel > MAX_UNORDERED ) {
>>>        unorderedChannel = MIN_UNORDERED;
>>>    }
>>>
>>>Luis | Tragnarion
>>>
>>>      
>>>
>
>_______________________________________________
>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/20060331/8903a1a9/attachment.htm