Sarah Jelinek wrote: > Hi Dave, > >> Sarah Jelinek wrote: >>> Hi All, >>> >>> Located at: >>> >>> http://www.opensolaris.org/os/project/caiman/auto_install/AI_Reqs_Final/ >>> >>> Are an updated set of AI project requirements. These have been >>> modified from the initial set sent out and reviewed a few months >>> back. Some requirements have been clarified, some have been removed >>> or reworded. Those that are specifically different are marked with a >>> **. The requirements were revisited based on the initial discussion >>> we had on this alias and the data we gathered during the prototyping >>> work we did for AI. >>> >>> The old, initial requirements list is here: >>> >>> http://www.opensolaris.org/os/project/caiman/auto_install/auto_requirements/ >>> >>> >>> >>> Please review and comment. The date for closure on comments for the >>> AI requirements is one week from tomorrow, Friday, June 20th, noon PT. >>> >> >> 3.2: it's not the language syntax, but the profile and rules that >> you're verifying. >> > Yes, will change. >> 4.1: "complete automation support" leaves much to the imagination and >> room for differing interpretations. I'd be more specific about what >> this means. >> > I think we have discussed this in subsequent emails. We are trying to > say that AI won't barf or do the wrong thing, and we won't offer a text > based installer to ask the remaining questions, if some of the required > data is missing. This is 'all automated' all the time. And, if critical > data is missing we will not perform the install. We will however default > on some things, like default locale, if we can. > >> 4.3: We're not providing upgrade in the interactive installer. Why do >> we provide it here? >> > I think we need to provide it here. The interactive installer targets a > different audience than AI does. Maybe it is the term 'upgrade' as > opposed to 'update'. I intend this to mean that we will allow for > 'updating' the system to the latest pkgs in the ips repo, via key,value > pair specification. I think enterprise customers would want this to > manage their servers. >
I'm unconvinced at this point. Why would anyone choose to take systems down, boot them from the net, and use AI when they can use a management agent (or really, just ssh) and do the "pkg image-update" without interrupting things. I feel here like we're attempting to recreate a dated concept which doesn't apply. >> 10.1: I don't see why the "IPS installation affects performance" item >> is here. Let's take a crack at some target performance soon, though, >> as we need real criteria to judge against. Performance from a server >> on a gigabit network, for example, should be better than installation >> from the CD. >> > Ok, fair enough. I will remove this. I do think that on all sides of the > install, cd with cpio, and ips, there are many, many factors that can > affect performance. Fast CDrom drive, slow network, vice versa, heavy > traffic on network, etc... > > To clarify the areas we need to look at with regard to performance: > -Client boot over network > -network speed and load > -ramdisk size > -ips pkg installations > A requirements document should set simple goals. These are design issues, not requirements. Dave
