On Mon, Aug 13, 2007 at 06:52:58PM +0200, Luis EG Ontanon wrote:
> I been thinking about the fact that Wireshark does not really need a
> MIB parser. It just needs to know OIDs, their name and their type.
>
> So I believe we should use a flat file that contains these three
> fields and provide the user with a separate program that actually
> reads the MIBs to maintain this file: like adding and removing
> modules. A file with basic oids should be provided to the user.
I'm not really happy with that: If I understand you correctly, you
intend to create an additional tool that will create an OID-to-Name
file which will then be used by wireshark. This will add another file
that will have to be installed, needs a manpage etc. Whenever someone
want's to add a new MIB they will need to recreate the file. I'd really
prefer a runtime solution for this, i.e. using libsmi directly from
libwireshark instead of an additional file. Is there a specific reason
you take a different approach?
> That way we could get rid of the NET-SNMP dependency, and *clean* a
> dissector that really needs it. I guess that more than 1/3rd (2/3rds
> if we take encryption out) of the handwritten code in
> packet-snmp-template.c can easily go. And trust me, no one will ever
> miss it, especially the maintainers.
No problem with that, on the contrary :-)
> The program to maintain the oids file (based on libsmi) I got it
> almost written (while "playing" with libsmi).
> The changes t get rid of the "extra code" in packet-snmp I'll commit
> in the next week or so if no one objects that.
See above.
ciao
Joerg
--
Joerg Mayer <jmayer@xxxxxxxxx>
We are stuck with technology when what we really want is just stuff that
works. Some say that should read Microsoft instead of technology.