On Mon, Jan 04, 2021 at 15:49:07 -0600, Ryan Gahagan wrote: > On Mon, Jan 4, 2021 at 8:24 AM Peter Krempa <pkrempa@xxxxxxxxxx> wrote: > > > On Tue, Dec 29, 2020 at 15:21:25 -0600, Ryan Gahagan wrote: > > > + <define name="diskSourceNetworkProtocolNFS"> > > > + <element name="source"> > > > + <interleave> > > > + <attribute name="protocol"> > > > + <choice> > > > + <value>nfs</value> > > > + </choice> > > > + </attribute> > > > + <attribute name="name"/> > > > + <ref name="diskSourceCommon"/> > > > + <ref name="diskSourceNetworkHost"/> > > > > This allows also 'port' and non TCP transports. It's okay to simplify > > the schema though and use the generic type. The code will need to reject > > those when we'll validate whether the disk is possible to represent for > > qemu. > > > > Just to be 100% clear, does this mean we can leave the > diskSourceNetworkProtocolNFS element definition as is and simply enforce > the port omission/tcp transport in the code? Or would you prefer that we > re-write the schema to use a custom variety of the diskSourceNetworkHost > which only supports tcp and has no port option? Yes, our schema doesn't cover all edge cases so we can use it here.
