Daan, This broke the KVM setups last week on master. I couldn't find your commit-id for this change scanning the git logs. Toshiaki-san was looking at CLOUDSTACK-3682 regarding this change [1] where he makes the scheme backwards compatible to work with KVM agents of the past. Could you please have a look at the changeset?
[1] https://reviews.apache.org/r/12985/ On Mon, Jul 01, 2013 at 10:28:57AM +0200, Daan Hoogland wrote: > The gain is that enums like BroadcastDomainType and IsolationType do not > have to check for different formats in the presented uri. It also makes it > more intiutive what the parts in the uri mean; vlan://<id> would inmply > that id is a hostname instead of a scheme specific identifier. > > I think I am reducing complexity, not introducing any. The present use of > uri for vlans is abuse of the construct, i don't think my proposed new use > is. > > > On Mon, Jul 1, 2013 at 10:17 AM, Hiroaki KAWAI > <ka...@stratosphere.co.jp>wrote: > > > Daan, I'm curious about what is the improvement of changing the > > vlan String format in URI? > > > > I'm -1 on pushing more complexity in URI, because that's > > abuse of URI class and sounds it's time to get rid of URI. > > > > > > > > (2013/07/01 16:47), Daan Hoogland wrote: > > > >> H, > >> > >> I've been trying to get a patch accepted and had some discussions to do it > >> in parts as well. I would like to refresh the first part: > >> > >> vlans are now referred as vlan://<id>. I would like to change this to > >> vlan:<id>. This will changee addressing the id as a scheme specific part > >> instead of as a host. As a result it will be easier to fix the code to use > >> Nicira NVP and other sdn networks whereever vlans are used now. Doing this > >> will result in a patch that is a subset of my earlier patch. It touches a > >> lot of core code and tests but has been tested thoroughly for VPC > >> gateways > >> with both vlans and Nicira NVP. > >> > >> regards, > >> Daan > >> > >> > > -- Prasanna., ------------------------ Powered by BigRock.com
