On 6/30/05, Pavel Cahyna <pcah8322@artax.karlin.mff.cuni.cz> wrote:
> On Thu, Jun 30, 2005 at 07:56:37PM +0200, Ivo Vachkov wrote:
> > > How? The hashes from the packet won't match the hashes in the table.
> > > Because in the table, there would be addrsss/mask entries which are n=
ot
> > > identical to the actual values in the packets.
> >
> > Masks will be used. Just like it works now.
>=20
> How will the hash be computed, then?

This is the part i'm currently working on :) I hope to give more
detailed information soon.

> > > Also, if you embed all the criteria in one routing table, how will yo=
u
> > > learn dynamic routing daemons (ospfd et al.) to add the
> > > source/protocol/TOS fields you need to the transient entries that the=
y
> > > create in the FIB? Because now routing daemons know only about the
> > > dest-address criterion.
> >
> > This is Quagga/Zebra developers problem :) Btw - since they support
> > Linux multiple routing tables it should be no harm to support NetBSD's
> > way ... it's just another #ifdef :)
>=20
> But how they should it? The Linux way with multiple routing tables seem
> straightforward to support, but I don't see how the OSPFD configuration
> would look with your scheme.

I think you're wrong here. The Linux way is just a hack to integrate
Policy Routing mechanism into iptables/iproute2/kernel.

> > > As the daemons (at least zebra/quagga) are modeled according to Cisco=
, it
> > > would be interesting to know how Cisco does it.
> >
> > Hmm, I'm not aware Zebra/Quagga to do any policy routing ...
>=20
> And Cisco?

How's Cisco Policy Routing related to NetBSD's ??? BTW Cisco uses
route maps to impelement it (correct me if I'm wrong), but after all
it's a whole different hardware/software arch under this. You have
integrated RPs, PBR, packet classifiers, filters, ASICs, backplane,
etc.

> Bye     Pavel
>=20


--=20
"UNIX is basically a simple operating system, but you have to be a
genius to understand the simplicity." Dennis Ritchie