On Tue, May 15, 2012 at 08:30:48AM -0700, Rob Sherwood wrote: > On Mon, May 14, 2012 at 9:23 PM, Justin Pettit <jpet...@nicira.com> wrote: > > Looks good to me. > > > > I wasn't volunteering that we'd change the default. There was a suggestion > > that OpenFlow 1.0.1 forbid standalone support. I said that we would most > > likely not follow that, since not only do we support that mode, it's our > > default. I suggested we might be open to changing the default, but I'd > > check to see what other's thought. > > Just fwiw, I too think it's a bad idea to forbid standalone support, > which is particularly useful for a vswitch. > > Per the conversation with Ben, my concern is the default being "melt > down the network if the switch has topology loops". There are a > number of possible solution to this including (1) make fail-secure the > default, (2) enable STP by default, or (3), the thing that I think > makes most sense, is to encourage hardware folks (who much more > commonly have networks with loops) porting OVS to consider fail-secure > as their default and leave the vswitch-targeted software (which rarely > has more than one uplink) fail-standalone. Ben's documentation change > is IMHO a great step in that direction; I would only consider adding a > note to the effect of "if you're porting OVS to a hardware box, this > may especially apply to you". > > Does that make sense?
It makes sense. I sent out a patch. Please review it, if you have a few minutes. _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
