Normally RSS served by bookmark services like http://del.icio.us do not contain all the links annotated by that user, but only the most recent ones. This is a crucial property of RSS syndication, which ensures that RSS servers serve only as the "sliding window" providing a glimpse to the most recent links annotated by that user. Nevertheless, sometimes we want to get access to ALL the links and the tags annotated by some user.
Write a software, which carefully (i.e. very slowly and with fake "User-Agent" headers, since there is a file http://del.icio.us/robots.txt forbidding most robots) crawls over all the links annotated by some user and concatenates the RSS feed, which has all the links, but otherwise its structure is identical to the RSS feed produced for the most recent links only.
If you wish to obey the restrictions from the robots.txt and do not want to risk any blocking action on behalf of the owners of the del.icio.us server, consider using a half-automated client (e.g. by running a plugin on top of Firefox, which asks you to configure the screen-scraper script and press a few buttons) rather than using a fully automated Java client, which is not a browser.
Third option would be to investigate the del.icio.us programmable interfaces, if any, and to use them in assembling the RSS file. At this point it is not clear, whether such an option exists (i.e. if such interfaces can be more helpful than direct screen scraping).