On Wed, Jun 5, 2013 at 8:19 PM, Matthew Flatt <mfl...@cs.utah.edu> wrote: > > * The details of the repository organization (including where to split > repositories) should be different. > > As described in next section of this message, the experimental > repository represents a revised proposal, but there's plenty of room > for further refinement.
Matthew asked me to try a less fine-grained organization of packages. My initial attempt at this is here: https://github.com/samth/racket/tree/pkg However, I don't think this is an improvement (also, it doesn't fully work). In particular, the dependency cliques are very large, and basically obviate the usefulness of a lot of the splits. There are some small packages, but basically everything depends on everything else. For example, Typed Racket brings in all of the gui libraries, all of the documentation, and the future visualizer. For me, this has been a useful exercise, and I now understand the constraints somewhat better. I think we have, roughly, two options: 1. Something like the split Matthew's tree proposes. In fact, I think we need to split some things further, so that `gui-lib` doesn't depend on scribble-related things. 2. Something much, much more coarse-grained, such as the current split between the 'textual', 'graphical', 'drracket', and 'full' distributions. Note that even these don't really make sense because of documentation build dependencies. I think that 1 is the right choice. I also think that continuing to develop in separate branches as proposals is a mistake. It's very hard to understand what's going on in the `pkg` version of the tree without using it -- I certainly didn't. it's also very hard to construct working trees in this fashion without anyone using the code. If we're going to make this transition soon, we should do it now, and then reorganize packages as necessary. Sam _________________________ Racket Developers list: http://lists.racket-lang.org/dev
