https://bugs.wireshark.org/bugzilla/show_bug.cgi?id=15018
--- Comment #13 from Guy Harris <[email protected]> ---
The whole endpoint thing is more fragile than I thought.
If a given protocol wants to keep track of information for a given "session",
where one of the protocols atop which that protocol runs defines a "session",
it wants to find *that* "session".
If we say a "session" corresponds to a conversation, there might be *more than
one* "session" to which a packet belongs. Consider, for example, TCP or UDP
running over IP running over an ISDN B-channel, where there might be a
conversation corresponding to the ISDN endpoint for the channel number *and* a
conversation for the IP addresses and TCP/UDP ports.
If the protocol runs atop TCP or UDP, it presumably wants *that* conversation,
not the ISDN conversation. If it can run atop TCP *and* UDP, it would be more
convenient if it could just call a routine that takes pinfo as an argument and
looks for the appropriate conversation - *even if there happens to be an
endpoint at a lower level*, so it should *not* search based on the endpoint.
In addition, the TIPC dissector used to set the source port for packets to its
port number, with a port type of PT_TIPC. It now sets an endpoint.
*If* tha port number was used to create a conversation, that might make sense.
*If*, however, the port number is used to set the "source port" column, that
won't work.
Perhaps
1) we need to have stacks of both information used for the columns and
information used to construct endpoints, so that we don't have to worry about
one dissector trashing information that might be used by a lower-level
dissector;
2) we need to have a way for a given dissector to search down the stack for
address/port information and for endpoint information, either getting the
topmost item or, if necessary, items for a specific protocol that might be
below the topmost item.
--
You are receiving this mail because:
You are watching all bug changes.
___________________________________________________________________________
Sent via: Wireshark-bugs mailing list <[email protected]>
Archives: https://www.wireshark.org/lists/wireshark-bugs
Unsubscribe: https://www.wireshark.org/mailman/options/wireshark-bugs
mailto:[email protected]?subject=unsubscribe
