On Tue, May 08, 2012 at 10:16:22AM -0700, Justin Pettit wrote:
> On May 7, 2012, at 5:08 PM, Ben Pfaff wrote:
> 
> > On Mon, May 07, 2012 at 11:56:52AM -0700, Justin Pettit wrote:
> >> Not all ports may fit in a Features Reply, so use the new port
> >> description stat message instead for looking up ports.
> >> 
> >> Signed-off-by: Justin Pettit <[email protected]>
> > 
> > Unless I'm misreading this (which is possible) it only works with a
> > switch that supports OFPST_PORT_DESC.  I don't think we should rely on
> > that, since no deployed OVS or other OpenFlow 1.[012] switch
> > implements it.  I think we should use OFPST_PORT_DESC as a fallback in
> > the case where the switch returns the maximum number of ports.
> 
> The thought occurred to me, too, but I didn't know how often people
> are not pairing ovs-ofctl with the locally built ovs-vswitchd.  But
> you're correct, that it's better to handle that so I sent out a
> revised version of the patch.

My own intention has always been that ovs-ofctl should work with any
OpenFlow switch, not just with OVS, and that it should work with other
versions of OVS too.  I've made a note to document that.

I agree that this is not the most common use.

> > As long as I'm looking at it, though, you can drop the use of n_ports
> > in fetch_ofputil_phy_port().  ofputil_pull_phy_port() will return EOF
> > when it's run out of ports to iterate. 
> 
> Man, you just don't let me get away with any laziness.  It's cleaned
> up in the revised patch.

Thanks.
_______________________________________________
dev mailing list
[email protected]
http://openvswitch.org/mailman/listinfo/dev

Reply via email to