Huge thanks to our Platinum Members Endace and LiveAction,
and our Silver Member Veeam, for supporting the Wireshark Foundation and project.

Ethereal-dev: Re: [Ethereal-dev] RFC: Feature candidate: Generic "Decode As..."

Note: This archive is from the project's previous web site, ethereal.com. This list is no longer active.

From: Guy Harris <guy@xxxxxxxxxxxx>
Date: Tue, 1 Jul 2003 15:53:53 -0700

On Tuesday, July 1, 2003, at 2:32AM, Biot Olivier wrote:

How about providing more generic "decode as" functionality by mapping a
protocol to a protocol field? We already have udp.port and tcp.port for the current "decode as", but how can I map e.g., "smpp.udh.ports.src == 12345"
to "wap-wsp"?

Mapping *arbitrary* protocol fields requires that the protocol tree be built; we don't always build it, for example, when initially reading the packets in, as doing so costs more CPU time.

Mapping arbitrary *dissector tables* is easier - that's what "udp.port" and "tcp.port" are in the "decode as" code, they're not protocol fields. (They happen to have the same names as some protocol fields, but they're not the same thing.) We already support that in Tethereal, with Lionel Ains' recent changes; you could, for example, match "smpp.udh.port == 12345" to "wap-wsp" in Tethereal with "-d smpp.udh.port==12345,wap-wsp".

That command-line flag is currently not supported in Ethereal, but that'd probably be a simple matter of copying the code over from Tethereal (and putting as much as possible in common code shared by Ethereal and Tethereal).

It might be a bit more work, but it might be possible to support that in the GUI for "Decode As" as well. One way to do it might be to attach to the "packet_info" structure a stack of dissector table match information, which would be maintained by "dissector_try_port()", with each entry in the stack supplying:

	the dissector table in which the match occurred;

	the port number that matched.

The GUI would change somewhat, as, instead of offering "Link", "Network", and "Transport" tabs, it would have to offer each of the dissector table/port pairs, along with a list of dissectors that can be used with that table. It also wouldn't allow specifying the source *or* destination port or both for TCP or UDP unless we, for example, had a "dissector_try_ports()" routine, taking two ports and checking both (in the same way "decode_tcp_ports()" and "decode_udp_ports()" do).