On Jul 4, 2014, at 1:36 AM, Peter Wu <peter@xxxxxxxxxxxxx> wrote:
> Should reassembly be enabled in as many TCP protocols as possible?
Yes.
> That is, is it desirable that TCP-based protocols use dissect_tcp_pdus with
> reassembly set to true?
That is a different question.
TCP-based protocols *where the data stream is broken up into PDUs, and the size of a PDU is always >= N, for some value of N > 0, and where the length of the PDU can always be calculated from the contents of the first N bytes of the protocol*, should use tcp_dissect_pdus with the reassembly parameter either hardwired to TRUE or passed as a user-configurable parameter that defaults to TRUE.
(That could be simplified to "should TCP-based protocols use tcp_dissect_pdus, *if possible*, with a default setting of "do reassembly?", to which the answer is "yes".)
However, not all TCP-based protocols are like that; HTTP, for example, isn't, and it uses different helper routines (see epan/req_resp_hdrs.[ch]).
