cvs commit: src/sbin/ifconfig ifconfig.c src/sys/net if.c if.h

[Bruce M Simpson]

Hi,

I have to admit to a certain amount of culpability here, as I wasn't really
thinking through all of the implications of Brooks' change when I agreed 'yes,
having an epoch timer which makes us less susceptible to roll-over where
struct ifnet and SNMP are concerned, is a good idea'.

Refactoring the PF_ROUTE ABI is probably a task for the 6-CURRENT lifetime.
If done right, we could MT5 it as we go along.

On Thu, Sep 02, 2004 at 11:36:44AM -0700, Peter Wemm wrote:
> BTW, in case nobody noticed.. nothing sets the routing socket abi 
> version number, and nothing checks it.  No versions are even defined.  
> Well.. version 0 is implicit I guess. :-)

Yeah. This is a pain.  There are a number of sticky things about routing
socket messages as they are right now.

This probably needs to happen using a different protocol family to begin
with, with the old PF_ROUTE kept on board as a shim; notice that PF_KEY_V2
is explicitly numbered as such, and has much in common with how routing
sockets and messages are dealt with.

A Tag/Length/Value scheme might be more appropriate; things like dealing
with netmasks are very hairy indeed.

This is not so much on my TODO list as my WISHLIST right now, XORP could
certainly benefit.

BMS
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 167 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/cvs-src/attachments/20040902/88b948ca/attachment.bin