On 07/10/2013 12:04 PM, Daniel P. Berrange wrote: > On Wed, Jul 10, 2013 at 11:51:42AM -0400, John Ferlan wrote: >> On 07/10/2013 10:49 AM, Daniel P. Berrange wrote: >>> On Tue, Jul 09, 2013 at 03:10:46PM -0400, John Ferlan wrote: ...<snip> >> >> To be sure we're on the same page, the storage_conf XML then becomes >> >> <auth username='someuser'> >> <secret type='[ceph|iscsi]' [usage='mypassed'|uuid='someuuid']/> >> </auth> > > Actually I think the schema is: > > <auth username='someuser' type='[ceph|chap]'> > <secret [usage='mypassed'|uuid='someuuid']/> > </auth> > >> That is there'd be no reason for 'type' in the XML nor 'password'. The >> 'login' goes away and the 'username' becomes required. > > I think you still want to keep 'type' in the XML, since if iSCSI adds > a different auth mechanism that isn't 'chap', then we have the ability > to represent that using a new type. > The <secret> XML has an existing 'type' which is 'ceph', 'iscsi' (eg, CHAP), or 'volume' and I just figured that since the <auth> XML requires 'username' and the <secret> subelement with a mandatory 'type' attribute, then that would cover the needs. See: http://libvirt.org/formatdomain.html#elementsDisks I'm not objecting to using the 'type' in the <auth> XML, just noting that's it's a duplication of an attribute, although I suppose if some authentication mechanism was added in the future that didn't use the secret element, then having a type present make things easier. John -- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list