Wireshark · Ethereal-dev: Re: [Ethereal-dev] resolving of well-known mac-addresses

[Guy Harris <gharris@xxxxxxxxx>]

On Wed, Aug 28, 2002 at 11:41:25AM +0200, Hannes Gredler wrote:
> IMHO it does make sense that these _protocol-specific_ [thats the important
> keyword here] MAC-addresses are registered by the dissector;

We also ship, and install, a bunch of XML files used by the Diameter
dissector, supplying protocol-specific information; there is *some*
Diameter dictionary information hard-coded into the dissector, but
that's just a fallback if your Ethereal isn't built with the XML
library.

As such, it doesn't bother me to put protocol-specific information, such
as multicast MAC addresses for particular protocols, into a
configuration file rather than hard-coding them into a dissector.

The configuration-file scheme also seems to me to be more extensible -
you don't have to keep tweaking dissectors to add new well-known MAC
addresses, you just add to the configuration file.

>   sure we could start messing up with files, however
>   these well-known MACs are specified in ISO10589 [10 years old spec]
>   and it is unlikely that they do change in the near future;

Yes, but that doesn't mean that there aren't other well-known MAC
addresses that we've missed - and it'd be nice if non-programmers, or
people with binary-only installations, could add the ones we've
forgotten (and send us patches to update the list).

> so i am in favour of hardcoding them into the protocol; the ethers files
> should be used just for site-local MAC address definitions;

"/etc/ethers", on UNIX, could be used for that.

We'd want to have a separate file for universal MAC address definitions
(although, unfortunately, right now, if you don't install stuff under
"/usr/local" or some "/opt" directory or, that stuff would be in "/etc"
- that is arguably a bug that should be fixed).