I'm not sure just how feasible or reliable all these ratings ideas might be.
Beauty: "One person's trash is another person's treasure" is a sign/epithet often seen or heard with regard to antiques and collectibles, but it might equally apply to music. A few months ago, I sang in a concert billed as "An American Choral Sampler" in which there were pieces by Billings, Bernstein, Copland, Randall Thompson, Eric Whitacre, and others. I recall that earlier working titles for the program were something like "American Choral Classics" or "Treasures of American Choral Music" ... It turns out that "sampler" was indeed a better choice than "classics" or "treasures" for such a concert. For example:
* One member, a friend and singer whom I greatly respect, thought the Randall Thompson "Peaceable Kingdom" (we sang the movements "Say ye to the righteous", "The paper reeds by the brook", and "Ye shall have a song") was simply awful, both as something to sing and as something to listen to, even though Thompson is highly regarded and often considered the dean of 20th century American choral composers. On the other hand, having sung the work several times, I regard the work (and these selecta in particular) very highly. My friend is gnashing his teeth over Thompson's "Last words of David" which we are singing with a different group in concerts this summer, and I am delighted to revisit this work and sing it once more.
* The "Sanctus" from Bernstein's "Mass" had widely mixed views amongst the singers, perhaps because (like so many Bernstein works) this work is as much musical theater as it is music, even though it's called music.
* "The promise of living" is the concluding chorus from Act I of Copland's opera "The Tender Land" -- an opera commissioned by Rogers and Hammerstein for the old NBC Opera Workshop and intended for television production, but it was rejected and eventually reworked for the stage. While the opera clearly has its weaknesses and has never been a hit, this chorus is sometimes sung in concert (as we did). Reactions ranged from beautiful to corny to not very good Copeland (I was in the latter category, although I highly value Copeland's music in general).
* "The Seal Lullaby" by Eric Whitacre is a gem that everyone seemed to like.
* "When Jesus wept" by William Billings (we used the CPDL edition by Rafael Ornes!!) is a round that paired well with the Robert Shaw/Alice Parker arrangement of "Wondrous Love" at the beginning of the concert, although I can think of Billings pieces that are, in some sense (to my taste), better.
In other words, I think that it is a very difficult task, as illustrated by the anecdotal information just described, to make an aesthetic determination of the beauty or worthiness of choral music, let alone an accurate one quantifiable by a number (or even a number paired with a standard deviation, which might at least give some indication of the range of "votes"), although I think such an assessment is probably not approprate to 1-dimensional quantification ... the Copeland mentioned above might have been evaluated as "corny but nice" or the Thompson as "nice but old-fashioned".
Popularity or page view/score download counts: Before CPDL had to go to a distributed ChoralWiki (with one Contributor wiki and two Visitor wikis) because of traffic, we had a reasonably reliable measure of a score's popularity via the hit counter for each page ("This page has been viewed xxx times"), retrievable in a Special:Popular scores page. Of course, these counts only reflected the total number of times pages had been viewed, so that most recently created pages were, necessarily, not as popular as pages created a long time ago. Furthermore, with the advent of the distributed wiki, these counts do not take into account views on the Visitor wikis (because they are replaced each day by the previous day's Contributor wiki data). So far, we have not been able to address this problem of counting views from the Visitor wikis and adding them to the counts on the Contributor wiki, perhaps because there might be considerable overhead/difficulty in implementing such a system efficiently.
Another count of such popularity might be to track the number of times an edition has been downloaded ... again something that we have not done (or figured out how to do without excessive overhead). If implemented, this would be an edition specific number, although it would not track downloads from externally hosted editions. So, once again, there is bias in the results obtainable.
Performance/sight-singing difficulty: This (performance difficulty) is something that high school music associations do (sometimes on a state-by-state basis) for choral music that might be sung at high school (or other) music competitions. It's also something I've never fully understood. That said, I wonder if we would obtain meaningful results, given the anecdotal nature of the way we might collect such information. Referring to my recent experience above: there were people that found the Bernstein "Sanctus" exceedingly difficult and others that thought it was rather easy or, at worst, average. The Shaw/Parker arrangement of "Wondrous Love" should probably have been easy, but putting together a good quality performance was rather more difficult than I would have imagined.
Slightly differently, a few years ago when my own "Creator of the stars of night" was being prepared for several concert performances, the conductor spent a great deal of time on it in an effort to get the best possible result ... something I had never thought would be necessary when I composed it ... happily, the effort was more than repaid by the reception it received.
Additionally, at the Contributor wiki, it is my most "popular" piece in terms of page views ... but that might be due, at least in part, to its being linked to from "Conditor alme siderum" (since the text is a traditional English translation of the Latin hymn). For that matter, I can only think that my second most popular work, "Veni, veni, Emmanuel", holds that position because of its title, since its difficulty (as reported to me) is definitely not in the "easy" category, and whether it is beautiful isn't for me to say.
Sight-singing difficulty is, I suspect, more difficult to appraise than performance difficulty. Such an appraisal seems to depend upon the sight-singing abilities of the rest of the group with which one sings. If all (or nearly all) singers are used to sight-singing, then I've found nearly everything is much easier than if the group has widely varied or mostly minimal sight-singing capabilities. My experience includes a semi-professional early music group (that ranges from 10-18 voices) that never rehearses with keyboard unless a work has accompaniment (so that everyone is "sight-singing" from the outset, even though there are singers better at this than others), two unauditioned choruses (of about 50-60 voices and 125-150 voices, respectively) that present concert performances (including the smaller group that presented the above-mentioned "American Choral Sampler"), an auditioned chorus (of about 45-50 voices), and several choirs of widely varying abilities.
Implementation and reliability: I've already alluded to the reliability of such results above, and I'm sure that these are tied to the way such assessment features might be implemented: download counts, versus view counts, versus "poll" style results whereby users/visitors contribute "feedback" assessments (by actually participating, which leads to various well-known inaccuracies for interpretation, for example results coming from "lovers" and "haters" or "squeaky wheels get the grease" contributions). Perhaps putting up a "feedback" poll would not be too difficult to do, since there is a MediaWiki extension for it (that seems, at least at present, to have some bugs), but it might take considerable tweaking to get it "right" ... Moreover, with over 11,000 score pages, will there ever be sufficient feedback to make any real sense out of the results?
To my way of thinking, I think that arriving at a workable view count that combines results from both the Contributor and the Visitor wikis (which can be refined to include views over a fixed recent period, as opposed to the cumulative counters available to the wiki implementation) for "popularity" is a valid statistic that might be useful to users. Secondly, I think that a download counter for individual scores might be useful, at least for scores where the PDF file resides at CPDL ... but implementation and maintenance of such a counter together with overhead are issues that would have to be addressed.
Chuck