On Mon, Sep 01, 2014 at 09:39:35 +0200, Jiri Denemark wrote:
On Sat, Aug 30, 2014 at 22:02:18 -0600, Eric Blake wrote:
> Took me longer than I wanted to get v3 posted. There's lots of
> new patches in this version, based on feedback on v2.
>
> Among other things, the bandwidth of virDomainBlockCopy is in
> bytes/s, and all the remaining interfaces are updated to use a
> flags value to decide whether to use bytes/s instead of the
> back-compat MiB/s.
>
> I really want patch 1 in 1.2.8; pasth 2-17 would be nice but
> it may be better to delay to later, and I still don't have
> virsh updated to cover the changed proposed in patch 18 so that
> one should probably wait.
Patch 1 is easy and safe to be included in 1.2.8 but I think it's way
too late for the rest. However, since the implementation will be missing
from 1.2.8, I think the best solution would be to just revert the patch
adding this public API and let it slip to the next release.
On the other side, we already did a mistake and pushed API when its
implementation was not ready yet and given that the API doesn't seem to
need any changes (except for the tiny one in 1/18), there's no strong
reason to revert it (except of for it being unsupported by all drivers).
It would be nice to at least have the remote driver implementation in
place but it's probably too late for that too.
Jirka