> Isn't the solution to have 3 levels: 'testing', > 'probation' & 'stable' ? > 'Testing' would be literally that, asking for > feedback from users; > 'probation' wb already tested for a defined period > -- say 30 days --
How about a crazier idea: Each package has a stability rating from 0-99 per architecture. 0 means totally untested/unstable and 99 means rock solid/no bugs. (0-33~unstable, 34-66~testing, 67-99~stable) Each new package starts at 50. Whenever a user uses the package, he can then vote on it by giving +1 or -1 (on the website or through portage). As a package gains rating points, more and more users would be inclined to use it. A user could set the minimum stability rating that he wanted for all packages or on a per package basis. After the user does an emerge update, the system would check if an installed package is above the users minimum. If not, the system informs the user of the drop and asks if he wants the package to be removed or not. Only the maintainer would be able to actually modify the package itself. Of course, big status changes would happen soon after a maintainer modifies a package. Questions to think about: 1) Should modifications reset the status of a package to 50? 2) Should a plus/minus vote forward propogate to the packages that depend on it? Eg. if kdelibs gets a -1 vote, should amarok get a -1 vote too? 3) Should the votes backpropagate instead? Cheers, Ben "he who writes the code gets to choose his license, and nobody else gets to complain" - Linus Torvale In my honest option, it should read - "he who writes the code gets to choose his license, and everybody else complains." -- gentoo-user@gentoo.org mailing list
