Daniel P. Berrange wrote: > On Fri, Jul 25, 2008 at 12:01:15PM -0400, Cole Robinson wrote: >> Daniel P. Berrange wrote: >> >> The general workflow is as follows: >>>> ========================================= >>>> import virtinst.Storage.StoragePool as sp >>>> >>>> # This gives the appropriate class for the specified pool type >>>> pool_class = sp.get_pool_class(sp.TYPE_FOO) >>>> >>>> # Only required params are a conn/uri and name. Default formats >>>> # and target paths have default values, but source paths/ >>>> # devices and hostnames obviously have no sensible default, but >>>> # they still aren't required for object instantiation >>>> pool = pool_class(name="foo", uri="xen:///") >>> Should proably allow a existing connection to be passed in, otherwise >>> this class has to deal with authentication issues too. >>> >> I didn't show it in the example, but passing a connection is allowed. >> >> On a side note, we should add support for openAuth in virtinst, >> everything is just hardcoded to use libvirt.open at the moment. > > No, that's exactly why we shouldn't take a URI. openAuth requires user > interaction, so that's unavoidably application specific. We just need > a helper in virtinst/cli.py for the command line tools to use to create > their connection object, and virt-manager already knows how to do > authentication. So everything should just pass ina connection object > instead of URI. > > Daniel Ahh okay, that makes sense. I'll remove the option to pass a URI then, since we really shouldn't even encourage passing a URI and expecting the library to open it. - Cole _______________________________________________ et-mgmt-tools mailing list et-mgmt-tools@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/et-mgmt-tools