A collection of votes is termed a vote source. We support two types of vote sources in order to support the two major classes of queries we expect users to make of the collaborative system. The first type of vote source is a specially formatted vote file. The second type of source is a vote server program.
Vote files provide our system with identified votes. Much work has already been done to create filesystems that allow tight control over who can read or write a file. Votes written into a file can be guaranteed both to originate only from one of the people with write permission on the file, and to be read only by those with read permission on the file. The intention is that if an individual creates a vote file, other users can consult that vote source to determine how the individual evaluated an article.
The presence of these individualized vote sources makes it possible to support group and custom moderation as users can create filters of the form, ``Show me the articles that Jane Doe liked.'' The system can answer the query by consulting Jane Doe's vote file provided that Jane Doe has provided the user making the query with access to her vote file. Jane Doe can control which people have access to her opinions in the same way she currently controls who has access to her other files, and could even store her votes for articles in different groups into different vote files to refine further her control over others' access to her opinions.
Vote servers as a vote source provide access to the summarized evaluations of the entire Usenet. When an evaluation of an article is requested from a vote server, the returned response represents a summary of all the votes posted for that article that have been received by the vote server. The information contained in the vote server supports the exploratory searching described earlier as presumably the best articles in a newsgroup will have received the most votes. New users exploring a newsgroup can quickly locate the articles previous users of the newsgroup thought were most relevant. Based on these articles, the users can decide whether or not to continue searching the newsgroup for information.