On 01/06/2013 04:04 AM, Ray Racine wrote:
On installation issues:1) Yes on defaulting to an installation-wide install as default. 2) A non-UI build installation, "server" build capability (in a backend/engine/runtime not web sense), Image manipulation libraries etc are installed but full DrDr and associated tooling is not. (e.g. fast build on a "cloud" server straight from git with minimal core collection/library support). - Building Racket from source on a AWS small 64-bit server fails. It shouldn't (last time I tried) and it is a niggling inconvenience that it is so. 3) Many (even most) of the collections library are non-core and IMHO should be evicted out into separate github hosted projects with drop-dead simple installation via Planet2 on demand. I know this sounds harsh, but collects is over due for a healthy paring, if not a full blown gastric by-pass.
+1
- I think an independence of development/release cycles between Racket and a host of stuff in collections is healthier for Racket and those libraries as well. Ray On Sat, Jan 5, 2013 at 9:58 AM, Matthew Flatt <[email protected]> wrote:I have been thinking about how developers who build their own Racket are more likely to want installation-wide packages instead of user- and version-specific packages. This is particularly true for those of us who work from the git respository. Maybe the default for a build `configure'd without `--prefix' (or, more generally, for a non-Unix-style build) should be installation-wide package installs, while the default for our pre-built distributions should be user- and version-specific installs. The default mode would be determined by configuration information in the installation-specific package area, and something like `raco pkg config -i default-mode <mode>' could change an installation's default. Does this sound like a good idea?
_________________________ Racket Developers list: http://lists.racket-lang.org/dev
