Re: DragonFlyBSD not in compliance with RFC 1122

[Hiten Pandya <hmp@xxxxxxxxxxxxx>]

Matthew Dillon wrote:
> :The router in question is a Cisco and is dropping the packets from 
> :192.168.2.0/24 that arrive on 192.168.15.1 due to anti-spoofing rules. I 
> :was just hoping for an elegant solution. As I've mentioned I added a 
> :work-around with IPFW but I'm not pleased with it.
> :
> :I was quite surprised that FreeBSD doesn't support this even in 5.3. I 
> :would have thought that the Zebra problem alone would be reason enough 
> :to implement it. DragonFly is obviously affected too.
> :
> :(The Debian machines I administer would just require a "route add 
> :0.0.0.0 gw 192.168.2.1" to add the second default route.)
> :
> :It's a problem that has gone unresolved in FreeBSD for over 18 months so 
> :I guess it doesn't affect enough people to be a priority. I was just 
> :wondering if it was planned for the ongoing DragonFly routing overhaul.
> :
> :Regards
> :
> :Gary
>
>     Ah, I thought it would wind up being something like that.
>     
>     This is actually a problem that has gone unresolved for over 10 years.
>     I've wanted to support multiple routes to a target forever.
>
>     The issue is that the route table module is a morass of unreadable code.
>     Jeffrey Hsu has been cleaning it up, and eventually I believe we will
>     be able to support multiple generic routes.  If you happen to be a
>     programmer and want to take a shot at it (hint hint) the door is open!
>
> 					-Matt
> 					Matthew Dillon 
> 					<dillon@xxxxxxxxxxxxx>


	This has been due to negligence on behalf of the BSD community.
	
	Anyway, this is possible target, all we need to do is bring in
	the RADIX_MPATH (multi-path routing) support from KAME, which
	was done by Itojun.  The other way to approach this problem is
	to decouple ARP from routing entries, but this is a more long
	winded approach... we could argue semantics all day. :-)

	Multi-path routing is something that is required some protocols
	like SCTP for fault tolerence.

	I suggest we use the RADIX_MPATH work from KAME tree, and adapt
	it to our new routing code; shouldn't be *that* difficult, IMO.

	Hope that helps.

			-Hiten
			hmp@xxxxxxxxxxxxx