I've been looking into CEDET, but it seems that most of its features would appeal more to developpers working in statically typed languages, and I'm kind of getting cold feet from the amount of tinkering it seems to require.
As I work mainly with ruby and javascript, I'm wondering what kind of improvements it could bring when working with these interpreted, dynamically-typed languages, over a plain TAGS file ?
CEDET is indeed a bit large, and the tinkering comes from the fact that there are so many different options. Everyone seems to want something slightly different so there is often a need to choose to configure it to ones own tastes.
The "smart completion" feature that depends on lots of type information is just a corner of the over all tool. There are some decoration type tools available. The parsers that already exist for java script and ruby will enable some decorating options, like drawing lines over the top of function tags, or enabling stickyfunc mode, which shows the current function in the header line.
It does perform some tags like functionality, but the tags can also be used in ECB, (the Emacs Code Browser), speedbar, or some fancy jump-to-tag like prompts. It can be used for basic completion (of global symbols) or for idle-summary-mode (like eldoc for other languages.)
Many of the other CEDET features still require that someone who uses these languages a lot write support for them, such as the project management system, or the code generator parts.