On 04/23/2013 07:44 PM, Eric Blake wrote:
On 04/23/2013 05:32 AM, Martin Kletzander wrote:
> On 04/22/2013 04:43 PM, Laine Stump wrote:
>> On 04/22/2013 10:13 AM, Laine Stump wrote:
>>> Of course in this case, we *could* avoid the camelCase vs underscore
>>> entirely by just naming the attribute "queues" instead of numQueues
or
>>> num_queues (the fact that the attribute is a number makes it pretty
>>> obvious that it is the "number of queues").
>> And I notice that the patches for supporting multiple queues in
>> interfaces uses <driver queues='n'/>, so consistency would vote in
favor
>> of using the same thing for <controller>.
>>
> I'm not the '+1' kind of guy, but this makes way more sense in the
> driver element and I think we would definitely make a use of that in the
> future, even though it increases the complexity of XML. Since there's
> still time, should I try to change it to <driver queues... or is
> somebody else against (or already did) that? (Last time I wrongly
> assumed that sending the patch would speed up solving the question, so
> I'm rather asking now).
It still hasn't been done. Martin, can you jump in and do it? Laine,
speak up now if you were doing it as part of your vfio <driver> work,
although it didn't look like that to me.
I thought Osier had said that he would make the change...
Yep, he says that here:
https://www.redhat.com/archives/libvir-list/2013-April/msg01565.html