Yes, it seems that the TCP checksum is not the real cause the problem.
After looking deeper in the source and doing some more tests, I see an interoperability issue in PHEX and Jtella.
The Gnutella RFC 0.6 says in section 2.3:
"Servents SHOULD remove any GGEP blocks from Ping, Pong and Push
messages before sending those messages to hosts that have not
indicated GGEP support."
Here, PHEX has not seen that Jtella-0.8 doesn't support GGEP, but sends the GGEP extension anyway. This causes Jtella to discard the PONG message due to a wrong payload exceeding 14 bytes.
AFter modifying some code in JTella, the PONG message coming from PHEX is now accepted, but I guess that a correction should also been done in PHEX, to avoid sending GGEP extensions to servents that doo not support this feature.

______________
Ivan Lovric
ESECO SYSTEMS

I just added the code to drop the GGEP data from pongs to SVN. I have no tested if it works correct, can you please do this...
You can get the changes directly from the Phex SVN or sometimes tonight or tomorrow from our automated build system (in case its running correctly): Index of /snapshots

I only did this change because it was quite easy to integrate, though actually I don't believe it is necessary to still support outdated nodes that don't yet support GGEP extension...

Gregor

__________________

I did also tests with QueryHit messages. The problem is the same: there are GGEP extensions in the QUERY messages sent by PHEX to JTELLA that are not well understood by JTELLA. After some hard coding in JTELLA, the problem is solved, but that should also need some adjustments in PHEX code.

___________
Ivan Lovric
ESECO SYSTEMS