On Mon, 2005-10-24 at 13:43 -0700, Patrick Mansfield wrote: > On Fri, Oct 21, 2005 at 05:56:50PM -0400, Jeremy Katz wrote: > > On Fri, 2005-10-21 at 14:55 -0700, Patrick Mansfield wrote: > > > On Fri, Oct 21, 2005 at 05:22:34PM -0400, Jeremy Katz wrote: > > > > While it might be useful eventually, I don't think it's something to > > > > really be concerned with for now. There are plenty of perfectly > > > > reasonable ways to get access to installation trees now and I'm not sure > > > > that I really want to add support for more given the impact on testing > > > > that it causes :) > > > > > > It is not a new installation method, it is similar to adding a new HBA > > > driver / module. I hope you don't have to test installation via all the > > > existing SCSI HBA drivers in the initrd, you would go insane. :) > > > > The fact that iSCSI uses the SCSI layer is purely an implementation > > detail, IMHO... the fact that it requires a lot of configuration about > > the target makes it much more like a network install type and FTP vs > > HTTP vs NFS are different installation methods > > It might require setup unlike current SCSI drivers and HBA's, but that > does not mean we should not support installs from it. > > iSCSI should be treated like just another transport like IDE, SATA, > firewire, USB, SAS etc. I'd agree if it were just another transport. The unfortunate fact of the matter, though, is that it's not. Otherwise, there wouldn't be any work which needed doing to support installing to or from it :-) Jeremy
