On Thu, Sep 04, 2003 at 08:33:36AM +0000, Sid Sid wrote: > As I can see from dissector codes high level dissector has to know how much > bytes do it want to and then ask low level dissector.
You've been looking at dissectors for protocols running atop TCP. TCP has no packet boundaries, so the protocols atop TCP have to tell TCP how much data they need. Look at protocols running atop IP, for example, such as UDP - they don't have to tell IP how much data to reassemble, IP can do reassembly of fragmented IP datagrams all by itself. COTP is like IP (or SCTP) in this regard; it's not like TCP
