On 13.01.2015 17:57, Eric Blake wrote:
On 01/13/2015 09:26 AM, Laine Stump wrote:
>
> I do understand Michal's concern. In particular, the network hook script
> support was put in specifically to allow management apps to support
> bandwidth management on interface types that libvirt couldn't/didn't
> directly support. If I recall correctly, the idea is that, although
> libvirt does nothing with it, the bandwidth info will be in the XML
> that's passed to the hook script when the interface is brought up, and
> the hook script can do whatever it likes to setup the bandwidth.
>
> Or am I remembering incorrectly.
>
> If that functionality is important, then this patch can't go in. If not,
> then it's a reasonable thing to do.
In the past, we've had other places where users were actively relying on
unsupported XML being preserved; when we tightened the schema, a
complaint was raised that we broke their setups, but we countered that
their setup was relying on undefined behavior and that we explicitly
have the namespace-aware annotations that they can use instead to
guarantee that XML will be preserved and untouched by libvirt. I'm okay
with stating that any user relying on this has been relying on
unsupported behavior, but agree that it is nicer to warn the user rather
than just silently break things. Maybe Dan has better advice?
Fair enough, let me respin with VIR_WARN (or is VIR_INFO more appropriate?).
I've pushed the 1/2 btw since it's a code cleanup practically unrelated
to 2/2.
Michal