BondingShouldBeFree Open-source Network Aggregation & Redundancy Solution

Chester A. Unal chester.a.unal at arinc9.com
Tue Jun 23 08:21:45 PDT 2026


Sorry for the late response. Your question prompted me to give this a proper
study. The current behaviour of the BSBF bonding solution is to create a TCP
stream per UDP 4-tuple and keep it alive until the connection had been idle for
300 seconds. The 4-tuple is source address, source port, destination address,
and destination port.

I'm planning to switch to the XUDP feature which puts a 16-24 bytes header on
each UDP packet. It provides address and port of the responding host which
brings full-cone NAT support.

As we're doing UDP over TCP, we will be stuck with the operation design of TCP
until MP-QUIC rolls around.

With that said, should we block QUIC traffic with the "xudpProxyUDP443"
option? QUIC over TCP is also room for meltdown as it is another feedback
control protocol.

Should we put multiple UDP 4-tuples into a single TCP stream? If yes, how many?
The "xudpConcurrency" option allows us to do from 1 to 1024.

At the moment, I'm thinking of doing one TCP stream for every 1024 UDP 4-tuple
and allowing QUIC traffic.

Chester A.

On 21/06/2026 20:06, Jonas Lochmann wrote:
> This confirmed what I already know, but did not answer my open
> questions.
> 
> So TCP is simple: One TCP connection is converted to one MPTCP
> connection that is unboxed at the gateway (lets call the remote server
> this).
> 
> And UDP is sent over MPTCP too. Do you use one MPTCP connection per
> UDP packet or per "session"? How do you deal with packet loss that
> should be propagated to the UDP packets instead of blocking the queue?
> 
> Am Sun, Jun 21, 2026 at 05:06:03PM +0000, schrieb Chester A. Unal:
>> Yes!
>>
>> https://github.com/bondingshouldbefree/.github/blob/main/profile/documentation.md#technical-overview
>>
>> Chester A.
>>
>> On 21 June 2026 19:14:49 GMT+03:00, Jonas Lochmann <openwrt at jonaslochmann.de> wrote:
>>> Am Sat, May 23, 2026 at 06:46:33PM +0000, schrieb Chester A. Unal:
>>>> I'm going to write a technical summary on the project documentation page
>>>> soon.
>>>
>>> Is this available now?




More information about the openwrt-devel mailing list