[mmann78@xxxxxxxxxxxx]

Some more thoughts about enabling/disabling heuristic dissectors based on some of the comments in the Gerrit review as well as this thread.

 

1. I'm now leaning more towards heuristic dissectors having their own name so something "more intelligible" can be presented to the user.  Right now we have heuristic dissector tables that just link a dissection function with a protocol (and that's put in a named table).  The "protocol name" is not enough to easily identify the heuristic to a user (but is enough for our current "internals").  With some of the "popular" protocols (Ethernet,  IP, TCP, UDP), it is fairly easy to figure out what's going on by concatenating "protocol name" and "table name" (ex ADwin-Config/udp can be understood to be "ADwin-Config over UDP"), but some dissectors are not really protocols and it just ends up looking ugly in the list (ex RPCoRDMA/infinitiband.mad.cm.private).  The same can be said for some "dissectors" (that really aren't "protocols") that ended up on the other "Enabled Protocols" tab.  This can happen when "identifier based dissectors" aren't really protocols but fit nicely into the provided internal architecture.

 

 2. I'm currently using the GTK GUI because work was already partially done towards the goal of a tabbed dialog that enabled/disabled heuristic dissectors.  However, I'm more interested in setting up the "epan code" to ensure the pieces are in place so any (Qt) GUI is capable of presenting something to the user that isn't just the "raw internals".  Some of the suggestions that have been made sound good, but the pieces aren't currently in place to connect the dots (and some look to require A LOT of work in the "epan code").  I'm okay with the GTK GUI version being "barebones" and mostly mimicking the current "Enabled Protocols" tab.  Qt can find a way to "pretty it up", but we can have the functionality in the mean time.  I'm mostly looking to finalize the "disabled_heuristics" file format so that doesn't have to be reworked (which could be a PITA if someone started using it, even if its between nightly builds)

 

3. Not sure how to integrate "all" heuristics together.  Ideally preferences, Decode As, the heuristic tables and even "identifier based" tables would all factor into the Big Switch, but right now each serves it own purpose and can provide specific granularity to certain use cases (usually allowing a user to override a "default (dissection/dissector) behavior" Wireshark provides).  The current Gerrit patch is just a small step in the right direction.

 

 

 

-----Original Message-----
From: Guy Harris <guy@xxxxxxxxxxxx>
To: Developer support list for Wireshark <wireshark-dev@xxxxxxxxxxxxx>
Sent: Mon, Jul 6, 2015 3:12 am
Subject: Re: [Wireshark-dev] Enabling/disabling ANY heuristic dissector

On Jul 5, 2015, at 9:33 PM, Hadriel Kaplan <hadrielk@xxxxxxxxx> wrote:

> My
2 cents:
> 
>> On Jul 5, 2015, at 11:32 PM, Guy Harris <guy@xxxxxxxxxxxx>
wrote:
>> 
>> "Heuristic Protocol" or "Heuristic Dissector”?
> 
> While
“Dissector” makes more sense to me personally, do most users/IT-folks understand
what a “Dissector” is?

That's why I prefer "Protocol".  Let's not let too
much of the internals show through.

> I think a single table will be more
confusing since several protocols have heuristic dissectors for more than one
underlying transport/protocol type.  Of course we could just enable/disable a
protocol’s heuristics for all underlying transports as all-onf/off... but I’m
just sure someone will have some reasonable use case for enabling heuristics for
some protocol over TCP but not UDP or vice-versa, and then we’d be back to
creating a preference for that protocol to do so.

So what exactly is the use
case for disabling "identifier-based" protocols?

Avoiding buggy
dissectors?

Disabling a level of protocol in which you're uninterested, so
that, for example, the Info column reflects the highest protocol level in which
you *are* interested?

For both of those cases, that's a use case for a Big
Switch for the protocol that switches off *all* dissectors for the protocol,
"identifier-based" and heuristic, no matter what protocol it's running
atop.

The use case for some but not other underlying protocols would appear
to be "traffic atop protocol X is rarely if ever mis-identified as being for
protocol Z, so leave the heuristic on, but traffic atop protocol Y is often
mis-identified as being for protocol Z, so turn the heuristic off".  Would that
be better handled by, for example, a UI to allow the user to specify the order
in which heuristic checks are done, or something such  as that (and a
command-line option to do the same, so that this same functionality is available
in
TShark)?

___________________________________________________________________________
Sent
via:    Wireshark-dev mailing list <wireshark-dev@xxxxxxxxxxxxx>
Archives:   
https://www.wireshark.org/lists/wireshark-dev
Unsubscribe:
https://wireshark.org/mailman/options/wireshark-dev
            
mailto:wireshark-dev-request@xxxxxxxxxxxxx?subject=unsubscribe