John Plocher wrote: > Shawn Walker wrote: >> As Bart and others indicated before, we can't provide this information >> anyway until the imageplan has been evaluated. > > I think we have a conceptual problem here, then. This design > and/or implementation forces the user down an awkward path that is > likely to generate complaints.
I disagree. Most Windows-based software or other software I've worked with is the same way. It doesn't tell you that a reboot is needed or will be needed until the very end :-) The worst of course is those that tell you that one is needed when it really isn't. Now, I think we can do better than that. So that isn't an excuse to not provide it at all. But it does point out that this is a commonly accepted practice. > I have a specific use-case in mind. I'm thinking of a Windows Update or > MacOS Software Update like scenerio that pops up with a list of 2 or 3 > product-focused updates: > > New updates are available: > MySQL Update > GNOME Update (user logout/relogin is required) > VirtualBox bugfix > NetBeans service pack > ZFS security patch (reboot required) > > I would *expect* that each of these would come with an indication > in the GUI that tells me that, if I choose to install that one, a > reboot would be required. Waiting for the user to click (install > now) before informing them of the consequences is (IMO) going to be a > significant user dissatisfier. I'm not sure I agree for the reasons I outlined above. However, there is a tradeoff here if we have to indicate that the user needs to reboot. Specifically, an evaluation of the updates available has to be performed so that we can accurately tell you whether or not one is required and as you might imagine, that involves relatively expensive disk I/O, etc. As Brock mentioned, it is probably possible to "cheaply" tell you that a "reboot" (that term seems so outmoded now...) may be required for the update or install of a particular package by using a special flag on a package. It all depends on whether you believe the expense of an operation to determine whether a restart will actually be needed is less or more than the cost of simply confirming that they still want to do the update. > This may be a different use case than you are designing for - most > of the discussion and code I see today focuses on long lists of > packages, and not on shorter lists of "Products". (see the linear > list of thousands of packages on http://pkg.opensolaris.org/status) That list of thousands is going away. It's rather misleading anyway as we re-publish every package for every build. Cheers, -- Shawn Walker _______________________________________________ pkg-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
