[Xorp-users] Redistributing the system kernel route table into OSPF

[Frank E. Renwick]

Bruce,

I wholeheartedly agree with your assessment that the optimized approach when
redistributing between routing protocols (such as OLSR and OSPF) includes
instantiating both under a common routing daemon, like XORP or Quagga.
However, there is much that can be done if one only has the capability of
interchanging information using the kernel as a "middle man."  For example,
in Windows, I currently use OLSR coupled with OSPF from RRAS.  By defining
an ASBR, the Windows box is able to create Type 5 LSAs (albeit only External
Type 2) from entries in the kernel route table. (I was attracted to XORP
OSPF because it appears to be able to inject External Type 1 LSAs, which are
much preferred.)

As a network engineer who is not a software developer, I was after the "low
hanging fruit."  I was under the assumption that XORP routing protocols
(OSPF,BGP, etc.) would be able to interchange with the kernel route table,
without worrying about where the kernel entries originated.  I also assumed
that this mechanism would be easier to implement than a full route
redistribution between two routing protocols, one of which is not currently
part of XORP.  As it turns out, with the version of OLSR I am using, if I
had the ability to redistribute kernel routes into OSPF, I would be able to
provide full bi-directional route redistribution between an OLSR domain and
an OSPF domain.  Assuming that XORP offers External Type 1 redistribution, I
would be able to include multiple redistribution points in the network
without fear of routing loops, by "sharing" the metric across the
redistribution boundary.

All this said, I do not expect that XORP should develop a capability just to
meet my needs.  I appreciate the discussion, and I will meet with my team to
discuss the option of participating in the development of OLSR as a XORP
routing protocol.  

Frank

-----Original Message-----
From: xorp-users-admin@xorp.org [mailto:xorp-users-admin@xorp.org] On Behalf
Of Bruce M Simpson
Sent: Saturday, April 15, 2006 3:16 AM
To: Pavlin Radoslavov
Cc: Frank E. Renwick; xorp-users@xorp.org
Subject: Re: [Xorp-users] Redistributing the system kernel route table into
OSPF

On Sat, Apr 15, 2006 at 01:30:40AM +0100, Bruce M Simpson wrote:
> > Currently, the FEA doesn't support this for Windows. I am not even 
> > sure whether the Windows API supports asynchronous notifications to 
> > userland if the forwarding table is modified; Bruce M. Simpson would 
> > be the right person to answer that question.
> XORP currently does neither because Microsoft's internal APIs are 
> still in a state of change.

The problem with the original approach (run olsr outside of XORP) is that it
relies on the routing protocol(s) being able to co-exist within the kernel
forwarding table -- it isn't quite the same as trying to use the kernel FIB
as a RIB.

The best way to make headway with your original application would be to get
the OLSR implementation for XORP cleaned up and integrated into the tree.

This would mean that OLSR itself wouldn't need to know how to talk to the
kernel forwarding table, and the XORP RIB can be used for supplying the
kernel with the correct paths, making things platform independent.

It also means being able to use the XORP configuration interface for OLSR,
and not having to worry about system specific means of implementing timers,
etc, as well as the rest of XORP being aware of OLSR.

Then as the base support for various OSes improves, functionality such as
on-demand path lookup can be added; the FreeBSD FIB supports this with no
kernel modifications; the Linux in-kernel AODV implementation modifies the
kernel in order to obtain such notifications.

Regards,
BMS

_______________________________________________
Xorp-users mailing list
Xorp-users@xorp.org
http://mailman.ICSI.Berkeley.EDU/mailman/listinfo/xorp-users