On Wed, Mar 17, 2010 at 04:48:54PM -0400, Michael Gilbert wrote: > so, i probably should have stated that i'm actually > thinking about the problem from a hypothetical > perspective; not necessarily with the current behavior of > gdp in mind, but instead in terms of what i think would be > more ideal. > > for the cases where packages don't have core data to begin > with, the launch scripts could tell the user that the game > data is missing, and that they should use gdp to load it > to the right place. By launch scripts, do you mean the actual game binary itself? e.g. /usr/games/foo? If so, the game packages dependencies will have to omit the data packages, which means you could install the games without having the data present and have a 'non-functional package', or at least, a game you can't play until you do something else. If you intent to wrap all the game binaries with a script which alerted the user that the data was missing, you would need to support operating in both text and X environments and handle translations. > obviously, doing anything this drastically different would > be quite an undertaking, and i can't expect anyone else to > do that work. i've actually been considering implementing > an alternative to gdp for a while now, and i'm getting the > motivation to start implementing it now. It's always useful to have a real mock-up that you can look at to appreciate a concept and see how well it works. I'd encourage you (and others) to put something together! However, I do not think that two solutions should co-exist in the archive. So, if you work on an alternative solution, the games team should debate the merits of both approaches and decide on which one is the best. If that were not gdp, then we should withdraw gdp from the archive and transition to the new scheme. The fundamental principle under which gdp was written was that packages should usable if they are installable. Stemming from this principle is a lot of work on data packages and dependencies which is not a lot of fun. It sounds like essentially your alternative proposal is to challenge this principle. I think there should be a games team consensus as to whether or not this should hold. Personally I don't enjoy working on game-data-packager. I see it as a "necessary evil", as the do-stuff-in-postinst model that was adopted by the flash installer, rott and a few others is unacceptable to me. I've also long since outgrown enjoying shell scripting :) But it has been my hope for some time that I could get it to a state where it didn't need a lot of time spent on it so I could work on something more fulfilling. If consensus is that the principle is not worth upholding, I'd fully support that. -- Jon Dowland
Attachment:
signature.asc
Description: Digital signature