Responding to your points out of order, since that makes more sense to me right now ...
Since currently DIB claims to be backwards compatible, we just need to > leave master backwards compatible with Kilo and Liberty Ironic, which > means not deleting the bash ramdisk element. If Ironic wants to remove > the bash ramdisk support from master, then it ought to be able to do > so. Yes, we'd like to remove support (read: code) from Ironic for the bash ramdisk. It was deprecated in Liberty, and I'd like to remove it soon (no later than once Newton opens). > What if you removed the code from Ironic, but left the element in DIB, > with a note that it only works with stable/liberty and earlier > versions of Ironic? > Sure, except ... > > Could we then: > > gate master DIB changes on an Ironic stable/liberty job that uses the > bash ramdisk - this would catch any regressions in DIB that break the > bash ramdisk > Yup. We could do this. > gate master DIB changes on an Ironic master job - this is what > gate-tempest-dsvm-ironic-pxe_ssh-dib is already doing (I think). > This, we could not do. Once we remove the support for the bash ramdisk from ironic/master, we will not be able to test the "deploy-baremetal" element in dib/master against ironic/master. We will only be able to test DIB with the "ironic-agent" element against ironic/master. However, since some users of dib still rely on the bash ramdisk (eg, because they're using older versions of Ironic) we understand the need to keep that element supported within dib. > > Is that a valid option, and would it remove the desire for a stable > branch of DIB? > We currently say that DIB is backwards compatible and doesn't use > stable branches. If there's a desire to change that, I think that's > certainly open for discussion. But I don't think we're in a situtation > where it's preventing us from moving forward with removing the bash > ramdisk code from Ironic aiui, but I might be misunderstanding. I also > think that having a stable branch sends the message that master isn't > backwards compatible. If that's not the message, why do we need the > stable branch? > > We believe we need the stable branch because we believe we should test master-master for "ironic-agent" and stable-stable for "deploy-baremetal". On the other hand, we could test master-stable (dib-ironic) for the "deploy-baremetal" element. If we did that, then we don't need a stable branch of dib. Thoughts? --devananda
__________________________________________________________________________ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev