We currently have a story about installing an item which goes like this:
"Jane visits http://xyz.org/bundle.xo[l] and her bundle is installed."
While adequate for the present, this story fails to explain how Jane found her bundle and it fails to explain what happens next week when someone unrelated posts a new version of the bundle, when Jane wants to suggest a revision, when Jane wants to remove the bundle, etc.
Consequently, we believe that a better story is needed.
Some of the things that Jane wants to do from the beginning to the end of her experience with a bundle include:
- install the "most relevant" bundle or be able to show several probably-relevant bundles in response to simple-text queries.
- select a language-appropriate variant of a bundle.
- distinguish between a "stable" bundle, meaning a bundle that no-one can improve any further and an "unstable" bundle whose contents and packaging are in flux.
- remove an installed bundle.
- look for the existence of other variants or versions of a bundle.
- show related things, perhaps bundles, as defined by the content packager.
- check how compatible an XO is with a bundle, e.g. by examining the XO for required viewers and usefully responding to their absence (i.e. give a warning, give installation instructions, or actually acquire the needed software)
- give feedback on an installed bundle
Relationship to existing systems and processes
These requirements are close to those of traditional Linux package management systems like rpm and dpkg; we should carry them out through an existing system. The way that we intend to manage the process of packaging content appears to be a bit different, given our bundle system -- we will have many packages with incomplete data. We must encourage and provide for packages that are not software but which are data (projects and information to be interpreted by other Activities), or are static browsable bundles; these will still have maintainers and dependencies.