On Wed, 29 Jul 2009, Ethan Quach wrote: > Alok Aggarwal wrote: >> Hi Ethan, >> >> On Wed, 29 Jul 2009, Ethan Quach wrote: >> >>> Alok, >>> >>> 3.1 >>> >>> For the boot choices for Sparc, I would really rather see >>> us reserve "boot cdrom" to not automatically boot AI at >>> all. I know this is AI-specific media, but at some point, if >>> our medias merge, changing the user expectation of this >>> interface becomes extremely difficult. >> >> That is a good point. Perhaps "boot cdrom - prompt" would >> be better? > > IMO, an additional specifier to some base that already > indicates "automated install" would make sense. So > something like > > boot cdrom - install prompt > boot cdrom - install standard > > ~or perhaps single token ~ > > boot cdrom - install_prompt > boot cdrom - install_standard > >> >> What should the behavior be if a user uttered "boot cdrom"? > > I think the AI services shouldn't be enabled in this case. > Maybe the system just boots to a login prompt?
I'll discuss this further with Frank and come up with an updated list of command line arguments for SPARC. >>> 3.2 >>> >>> Where do network-based AI boot and media-based AI >>> boot converge? In other words, when an AI client boots, >>> where does the knowledge in differences in boot-style end? >> >> The two converge after the service to locate a manifest >> to use has found a manifest to be used. The auto-installer >> service will subsequently invoke the installer in the same >> way regardless of media type. > > There's not much of a description in 3.4, but the separated > "manifest-locater" service shouldn't have to be media-based > vs. networked-based boot aware either. Is this correct? > In general, I think the convergence should happen as > early as possible. I think to a certain extent, this "manifest-locater" service will need to know about the boot media because in the current implementation, whether service discovery (/usr/bin/ai_sd and friends) is kicked off is a function of the boot media. Do you see it differently? Alok
