[ENet-discuss] HUGE Packet Loss ~80%

Ben Johnson bencelot at gmail.com
Fri Aug 22 13:59:48 PDT 2008


Hey all,

I'm experiencing MASSIVE packet loss.. only about 2 in 10 packets are
getting through from the client to the server.
Server to client works just fine. But when I try sending unreliable packets
from the client to the server, only 20% or so go through.

The packets aren't even big.. For testing, all I'm doing is sending over a
unique number each packet, that increments in the form of a counter, and
printing it to the console in both applications (client and server).
On the client side, I see 1,2,3,4,5,6,7,8,9,10,11,12 etc...
but on the server all I'm recieving is: 3,11, etc..

The weird thing is.. if I do this back the other way (server to client), it
works fine.
Also, if I use ENET_PACKET_FLAG_RELIABLE when sending from the client to the
server, it works fine too. I'd have thought that if I were getting that much
packet loss,
that it would take ages for the reliable packets to send properly as well..


Funnily enough, I only just noticed this, despite having played my game many
times multiplayer. It's because I'm using the co-simulation design described
by Lee here:
http://lists.cubik.org/pipermail/enet-discuss/2008-March/000820.html

All the commands were being sent eventually (as they were stored in a
queue).. but as that queue was always about 5 timesteps deep, the movement
wasn't responsive at all.
I thought this was a bit strange, so did some testing, and discovered that
it's because only 20% of packets are going through! If I send these inputs
reliably, it works great,
although I don't want to, for the reasons outlined in the link above.

I've tried a whole bunch of channels, even giving it it's own unique
channel, but no change. I also tried sending the packets unsequenced.
Nothing works except making them reliable.

Any ideas at all what could be causing this?

Thanks,
Ben.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cubik.org/pipermail/enet-discuss/attachments/20080823/c6aeb6b0/attachment.htm>


More information about the ENet-discuss mailing list