Version compatibility is a possible partial solution but the problem is that we fail to do the install with unexpected or undefined results. However I don't want to talk about solutions just yet.
Sent from my iPhone On May 19, 2009, at 7:04 PM, Jack Schwartz <Jack.A.Schwartz at Sun.COM> wrote: > Hi Evan. > > I like the first part, but then it seems that the problem statement > goes out of scope. Version compatibility doesn't have to do with the > failure not being handled or logged gracefully. (Hopefully the > latter will be fixed as another part of the project.) > I don't think that version compatibility is the full scope but is more of a possible solution to parts of the problem. I believe that the error and information handling is in scope. > Version compatibility has only an indirect connection with insight > as to where or why the failure occurred (when versioning prints a > message that things might not work upon detecting a version > mismatch), but OK. > > If you want to keep the problem statement, grammatical corrections > follow: > > On 05/19/09 17:28, Evan Layton wrote: >> Problem Statement: >> >> * Currently installing a client where the booted image doesn't >> match the OS bits being installed, the installation will fail > ... the installation could fail ... The point here is even if we don't fail completely the install that results is somewhat undefined. Due to this "will fail in undefined ways" is acurate. > >> in unexpected and undefined ways. >> * Further the user is given no information on why we failed > Further, the user ... on why the installation failed ... >> and the failure is not handled or logged gracefully. > ...because the failure is not handled or logged gracefully. OK that that sounds a bit clearer. > >> * Additionally both the user and the installer application >> have no insight into where or why we failed. > Who is "we" here? Isn't the installer application the thing that > failed? Or at least the application should know how the installation > failed because it was running the installation? > Oops, yes that should have been "the install application" failed. Thanks for the comments!!! -evan > My $.02, > Jack >> >> I would like to get any input from Ethan, Karen, Jack, Frank and >> anyone else interested on the problem statement by 5pm PST on >> Wednesday. >> >> I've also added a wiki page for this information and added meeting >> notes as well as other information. >> http://opensolaris.org/os/project/caiman/CVERS/ >> >> Thanks! >> -evan >> >> >> More notes... >> >> The areas of possible incompatibilities: >> 1) Architecture, SPARC or X86/64 (for image or archive based >> install). >> 2) OS version the client is booted to >> 3) installer version >> 4) The type of install image we are using. For example Package >> based, archive(cpio, etc) based or image (replicated or recovery >> image) based. >> 5) OS version being installed >> 6) manifest and schema versioning to make sure they match. (This is >> part of Jack's XML parsing work) >> >> >
