On 1/29/19 8:49 AM, Andrea Bolognani wrote:
> On Wed, 2019-01-23 at 16:32 -0500, Cole Robinson wrote:
> [...]
>> +++ b/docs/schemas/domaincommon.rng
>> @@ -2499,6 +2499,15 @@
>> </element>
>> </optional>
>> </interleave>
>> + <optional>
>> + <attribute name="model">
>> + <choice>
>> + <value>virtio-9p</value>
>> + <value>virtio-9p-transitional</value>
>> + <value>virtio-9p-non-transitional</value>
>
> I thought there was rough consensus on having separate 'model'
> and 'protocol' attributes, with the former using the same values
> as other VirtIO devices, but looking through the archives I've
> found
>
>
https://www.redhat.com/archives/libvir-list/2019-January/msg00799.html
>
> where you said you were going this route for v2... Sorry I didn't
> notice earlier and thus didn't have a chance to yell :)
>
> I think being consistent with other devices is more important than,
> for lack of a better term, "marketing" virtio-fs.
>
> Moreover, management applications like virt-manager and Cockpit
> will probably present this as a single drop-down to users, so it
> hardly matters that it ultimately ends up being translated to two
> separate attributes and what the corresponding values are.
>
Okay I'll go with the protocol= syntax danpb suggested
Althought I guess the protocol= syntax is really only interesting once
we have another use case like usb-mtp or virtio-fs wired up. So I guess
I'll just do model=virtio|virtio-transitional|virtio-non-transitional
- Cole