[ENet-discuss] Channel Incoming Unreliable Commands List Management Question

Lee Salzman lsalzman at gmail.com
Wed Aug 1 09:46:16 PDT 2012


Note that if ANY fragment of an unreliable packet gets lost, the entire packet is gone, i.e. you should not be trying to send a packet with 200 fragments via unreliable fragment because there is almost a 0% chance you will ever receive the packet because packet loss of some form is guaranteed over that many UDP packets.

On 08/02/2012 02:12 AM, Alexander Dolgansky wrote:
> Hi Lee.
>
> I was quite excited to see the update so fast! Thanks.
>
> So, I've tried it out and now I can send messages that I previously
> was not able to send wirelessly using best-effort delivery method. The
> only snag now is that if I attempt to increase the message size (~200
> fragments), I can no longer re-assemble the received messages using
> either wireless and wired setup. The wired setup did work before with
> the same kind of message size, so I am not sure why it stopped working
> now (the wired setup did not change at all). Moreover, the increase in
> message size does result in a larger bandwidth usage so it appears
> that large number of packets do reach the intended target.
>
> I wonder if I am hitting the limits of what I can send unreliably but
> again it did work with the wired setup. Also, sending the same message
> back to back results in only the first message being successfully
> processed by ENet. Slowing down the sending rate for larger messages
> doesn't seem to help either so it almost feels like that once fragment
> count reaches certain value, messages begin to disappear continuously.
> This is just my speculation. I did notice that unreliable sequence
> number is only two bytes long but that will mean that a message will
> need to be partitioned into more than 65535 fragments to cause an
> overflow in the sequence number. I don't believe I am anywhere close
> to that number of fragments per message, so overflow is likely not the
> issue.
>
> I would like to pursue this issue a bit more to see if there is still
> an undiscovered ENet bug or if I've reached the limit of what I can
> send using the unreliable fragment option.
>
> Alexander.
> _______________________________________________
> ENet-discuss mailing list
> ENet-discuss at cubik.org
> http://lists.cubik.org/mailman/listinfo/enet-discuss
>



More information about the ENet-discuss mailing list