Future

Rewriting of the WWW frontend scripts in C/C++ has the highest priority. A perfect solution will be Apache module or incorporating module in some small and fast HTTP daemon.
URN support. Currently Squid has only very basic URN support. The solution will be the automatically transformation of the ftp request into URN resolution request.
In order to update list of the ftp servers custom parsers of the comp.* newsgroups and freshmeat are needed.
It will be nice to have a tiny daemon that will provide false ICP-HIT responses to the cache servers and generate pages with HTML responses. In such a way we will be able to significantly reduce MISS response time. It will also provide a hook to the cache servers without URN support.
Someone has to push authors of the browsers for inclusion of URN support.