On 05/08/2013 06:49 AM, Kevin Wolf wrote: > Am 29.04.2013 um 09:42 hat Stefan Hajnoczi geschrieben: >> @block-backup >>
> drive-backup would probably be a more consistent naming. We would then > still have block-backup for a future low-level command that doesn't > create everything by itself but takes an existing BlockDriverState (e.g. > created by blockdev-add). At least it would match why we named a command 'drive-mirror' instead of 'block-mirror'. Hmm, looking at qapi-schema.json, I wonder if we can rename 'BlockdevAction' to 'TransactionAction' as used in the @transaction command. It wouldn't change what is sent over the wire in JSON, and until we have full introspection, there is no visibility into the type name used. Changing the name now would let it be more generic to adding future transaction items that are not blockdev related. > > We should also make it transactionable from the beginning, as we don't > have schema introspection yet. This way we allow to assume that if the > standalone command exists, the transaction subcommand exists as well. Agreed - existence of a command at the same time the command is made transactionable serves as a nice substitute for not having full introspection into the 'BlockdevAction' union type, whereas if we introduce the command now but not transaction support until 1.7, life becomes tougher to know when it can be used where (although I HOPE we have introspection in 1.6). -- Eric Blake eblake redhat com +1-919-301-3266 Libvirt virtualization library http://libvirt.org
signature.asc
Description: OpenPGP digital signature