[Thincrust-devel] selinux security context

[David Huff]

David Lutterkort wrote:
> 
> Yes, they are very different formats; libvirt XML[1] is geared towards
> creating a VM on a specific host/hypervisor. The virt-image XML[2] is
> focussed on describing a VM image in a way that can be used to run that
> image on a variety of hypervisors.

David thanks for the clarification.  Every thing you mentioned is kinda 
what I was thinking.  It looks like the fundamental problem is that, the 
logical place for this kinda of functionality (Image object representing 
a virt-image described VM and the conversion to other formats, ie 
vmware) is in virt-convert.  However it looks like the stuff in virtinst 
is better written and easier to use.

> 
> The conversion from virt-image XML to VMWare should be in virt-convert
> (if it's not, that's where it should go)

It looks to me like this functionality is in both tools.

> 
> The virt-convert code for handling virt-image files does look a little
> strange .. rather than filling in fixed XML templates, it would be much
> cleaner to add toXML methods to the various classes in ImageParser so
> that you can just call Image.toXML and get an XML representation of the
> internal objects.

This is the path I was looking to go down.  Looks like it would easy to 
build an imageParser obj which I could have a toXML method then use the 
unWare class to build a vmware conf file.

Again the problem is that it seems that using virtinst to do this is the 
worg way architecturally to go about this.

Any comments on this, or am I totally over analysing this?


-D