On Jun 10, 2016, at 10:40 AM, Robert Cragie <robert.cragie@xxxxxxxxxxxxx> wrote:
> It's more like a tunnelling protocol e.g. L2TP where L2 runs atop a L4 protocol - which is not "conventional".
"Conventional" is irrelevant to Wireshark - from the standpoint of Wireshark, X runs atop Y if:
Y has a mechanism for handing its payload to other dissectors;
X's dissector is one of the ones to which it can hand its payload;
regardless of where X and Y "normally" happen to live in the protocol stack.
So, if protocol Y has a dissector table and thus also has support for "decode as", registering X's dissector with Y is done the same way regardless of whether the layering is "conventional" or not; "unconventional" must not affect how you think of setting up your dissector - or how you ask about how to set up your dissector.