Next Generation Ruby packages for Ubuntu

imo, I don't think Debian and Ubuntu should do this - it's not really their job. Distros should, for the most part, just be packaging upstream code and not maintaining an unofficial fork.

Though counter to this, RedHat basically maintain their own fork of the Linux kernel but it's a bit of a special case, and they do employ full time kernel developers to do it!

> Why isn't this work a part of an official ruby release?

nobu's lazy sweep collector is in trunk. Sokolov Yura's patches have been submitted; some of the merged, others introduce slight incompatibility issues.

Debian and Ubuntu shouldn't be in the business of trying to second-guess upstreams. They should be packaging things up as they're released.

They should build without much modification - it just needs a bit of effort to setup a repository and do the builds. If there is demand for it, we could look at doing it.

Rebuilding the source packages should be easy enough, just loop over the .dsc urls and run dget -x; sudo mk-build-deps -i; debuild; sudo debi. pbuilder and sbuild also work for cross-builds, but I didn't get the hang of cross-building entire package graphs.

They do. Check out https://gist.github.com/1688857

1.9.3 is stable and fine for deployment.

Already in production for months, no particular problems

1.8.7 remains popular, but is declining, and for good reason. While there are some areas where 1.9 branches gave up ground (I'm looking at you `require` [1]), there are other areas where it absolutely smokes older versions; like date formatting functions [2]. Fortunately, Rails processes run for a long time under most app servers, so the performance regression of `require` isn't a huge deal. It is, however, very common to format dates as part of output, which may happen on hundres or thousands of records. There are also some nice features of 1.9 branches that are useful in a practical sense. For example, hash order is presered in Hash under 1.9, so you can get the extra features of a hash, and avoid superfulous sorting (can result in big compute savings).

[1] https://github.com/bradland/require-benchmarks

[2] https://gist.github.com/1352997

1.9.3 is great for production. You give up a bit of memory savings compared to REE (since Ruby proper won't be COW-friendly until 2.0), but 1.9.3 in general is faster, has a lot more features, and is widely supported by gems and libraries now.

1.8.7 has been EOL'd, which is as good a reason as any to upgrade.

You are not even remotely close to being correct.

I just switched from 1.9.2 to 1.9.3 specifically because it cuts Rails startup time in half.