On Tue, Mar 15, 2016 at 02:14:42PM -0700, Guru Shetty wrote: > On 15 March 2016 at 11:32, Ben Pfaff <b...@ovn.org> wrote: > > > On Sun, Feb 28, 2016 at 10:33:24PM -0800, Gurucharan Shetty wrote: > > > Signed-off-by: Gurucharan Shetty <g...@ovn.org> > > > > > Thank you for looking through the series. I agree with all your comments so > far in the series and I will get them right in the non-RFC series. I have > one comment below. > > I'm uncomfortable with the form of the argument to ct_lb. It seems odd > > that it would be a string, since it is naturally a list of IP addresses > > and the OVN match/action language is well suited for lists of IP > > addresses. > > > > I went with the string as that was what needs to be given to the created > group string of the > form: > type=select,bucket=bucket_id=%u,weight:100,actions=ct(nat(dst=%s),commit,table=%d,zone=NXM_NX_REG5[0..15]) > ...... > > So if I understand you right, you want the ip reconverted again to string? > So the advantage with your approach is that if a wrong string is given, the > translation to IP addresses will catch it early?
I want the syntax of the OVN logical match/action language to make sense independent of their implementation. The logical matches and actions don't currently have anything where a string needs to be parsed into tokens to understand what's going on, and unless you're very familiar with the implementation then it doesn't make sense. I don't want whoever writes the OVN logical flows to need to be familiar with the implementation to understand. I also agree with the reason you suggest. _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
