On 2008-2-11, at 9:36, ext Stig Venaas wrote:
I can see at least 3 possible theoretical effects of delaying the first
packet.
1. First packet is delayed, packets are kept in order
2. First packet is delayed, subsequent packets may arrive first
3. All initial packets dropped
...
Well, this is still more or less speculation. I suppose what
should be done, is to study some common applications or
protocols and see what the effects would be. An obvious first
candidate would be DNS I think.
Exactly right. I've said this a few times in the past; data on how these different behaviors affect applications and transport protocols are essential before we can determine what's acceptable.
Although UDP-based apps (e.g., RTP) are important, what will happen depends a lot on the individual application. Initially, it would be easier to investigate how TCP/SCTP/DCCP react when 1, 2 or 3 happens, because their behavior is more tightly specified.
we need to look at the size of the buffers needed to store packets and the overhead introduced on the control plane of routers.
This can be an issue if there are too much packets to manage.
For example, for TCP, I'd expect that in case 1, the initially sampled RTT would be too high, causing a slower slow-start, in case 2, there'd be a few 500msec timeouts during the first few roundtrips due to the reorderings appearing as losses, and for 3, a 3-second timeout will kick in before the SYN will be retransmitted. But that's just a guess - I haven't measured/simulated anything, either.
Lars
--
word 'unsubscribe' in a single line as the message text body.