I've extracted out the code I was using to lazily load large BLOB/TEXT columns in one of my projects, tidied it up to take advantage of a refactoring in Rails 2.2.2 (cheers Koz) and improve its semantics, and released it as the columns_on_demand plugin.
After upgrading a server to Ubuntu Intrepid I found that the Rails site's layouts broke. The problem is this issue resulting from a ERB code change in Ruby 1.8.7: if you put a # Ruby comment in a one-line <% %> block, all the content up to the next %> that's on a subsequent line is ignored.
I've just published a tiny Ruby on Rails plugin that adds a script/dbdump command.
I've submitted another patch to Rails for a minor-but-annoying bug – if you use db:migrate:redo you'll find that it does migrate correctly but the schema dump afterwards doesn't work, so db:test:prepare clones the wrong structure (for example).
I've created Rails lighthouse ticket #1110 for another bug in the new 2.1.x association preloading code that was causing some has_manys to get loaded with two copies of each record in the collection (and less seriously, double SQL loads of all types of collections).
I've created Rails lighthouse ticket #1101 for a blatant 2.1.x bug breaking :including of associations that have :conditions hashes.
I'll be speaking tomorrow, Wednesday 2009-05-19, at the first meeting of the Christchurch on Rails group.
One of the great things about Ruby frameworks is that we use Ruby code to configure and initialize all the other Ruby code, so you don't need to mess about with endless XML configuration files and write wrapper classes just to run some little bit of arbitrary code to setup your application when the server starts up.