V2.01 uTP issues/bugs [4] (New header - backward compatibility dilemm)

[rafi]

- Fix: Fixed bug in uTP header field layout

After a few checks with uT 2.01 and 2.0 , it seems there can be a backward compatibility issue, causing clients using the new header to not communicate sometimes. Here is what I've noticed for different bt.transp_disposition values:

2.01 peer 2.0 seed result

---------- -------- -----

31 31 TCP (not consistent)

31 26 -

31 10 uTP

26 10 uTP

26 26 -

10 26 uTP

2.01 seed 2.0 peer result same as above except:

---------- -------- -----

10 26 -

"-" = no transfer/connection

My assumption is that with 1.85 the new header will not work at all (should be verified)

I suggest to set the default of version 2.01 to 31 (enabled). Optionally with the following settings:

26 for "bandwidth management" 'forced"

31 for "bandwidth management" 'enabled'

13 for "bandwidth management" 'passive/disabled'

(note: the "5"/disabled can be added too)

All with a similar UI as the "encryption".

The result being - that by default-uTP will used mostly between 2.01 peers (that have the most debugged version).

On 2.0 peers and below it will revert to TCP. Gradually - 2.01 peers # will increase and at the same time 2.01 will mature, keeping most of the clients with the older/buggier uTP versions - out of the loop.

[Firon]

This isn't a bug.

The new uTP header isn't implemented in 1.8.5, and 2.0's current implementation of it is broken (ditto for 2.0.1 older than 18408). We're making a new 2.0 to fix that today. 2.0 still uses the old header, so all connections received from 2.0 will work regardless.

Closing thread as non-issue.