OoO Pendant  le temps  de midi  du lundi 23  novembre 2009,  vers 12:56,
Jochen Friedrich <[email protected]> disait :

>> - As stated several times in the discussion leading to this change,
>> libsnmp-base doesn't necessarily need these MIBs to function; only the
>> command-line tools, and *possibly* other software using libsnmp-base
>> that wants names rather than numbers, do.  That would tend to lead to
>> a "Suggests" at most from libsnmp-base.  Prompting for installation
>> proves quite intrusive by comparison, for something most users don't
>> need.

> The problem is I don't really know which package only needs libsnmp-base
> or libsmi2-common and which package needs the MIBs.

libsmi2-common  contains  just  MIB.   Some packages  sets  an  explicit
dependency  on this  package (wireshark-common  and  python-pysnmp4) and
therefore should  suggests the future contrib package  that will replace
it.

> I'm really into moving the MIB installation to a package in contrib. This
> package should probably depend on libsmi2 as smistrip does a better job than
> the script included in the net-snmp package.

libsmi2-common also contains many more  MIB than the previous version of
libsnmp-base but this is a lot of files that should be checked for their
licenses. We could just package  a non-free package from libsmi2 sources
with those MIB  (provided we do the license check).  This would allow to
ship MIB  without downloading  anything. We should  just create  a dummy
package suggesting the  non-free one and containing some  note about the
necessity  to install  the  non-free package  to  get MIB.  Applications
requiring MIB may depend on  this dummy package or suggests the non-free
one.

However, since IETF MIB have already been cleaned (i.e modified) I think
that there is still some legal aspect in this.
-- 
 /* Thanks to Rob `CmdrTaco' Malda for not influencing this code in any
  * way.
  */
        2.4.3 linux/net/core/netfilter.c

Attachment: pgpJ9Yfh4Pnr3.pgp
Description: PGP signature

Reply via email to