CmdrTaco's Journal: Buglets and Bowls and Comment Scores

Super Bowl was pretty fun. I didn't see no stinkin' X-Men 2 commercial, but dammit it I'm so stoked for The Matrix.

Lots of stuff going on in the code lately, and I'm getting more than the usual amount of flame bitching about it. Since I'm back from New York (pretty busy week all things considered, but I had a better time than I was expecting) I actually have some time to address a few things. This weekend has been loaded with heaping flames (relating to someone who spoofed my email on their web page) and a number of people bitching about me not staying on top of alerting people to changes on Slashdot.

First up, its worth noting that the last couple weeks have been somewhat hellish. Over Christmas we had some robot attacks, and some pretty serious hardware and software issues. Because of these things, and of course the shortened holiday schedules, we weren't able to do our usual wed. night code refresh. For those of you who don't pay attention, jamie usually syncs Slashdot to the latest tagged release on CVS every wed. Normally our code syncs are a few hundred lines of random stuff, but since this one was nearly a month worth of code and bug fixes held back because of the serious performance issues, this one was a doozy.

So there have been lots of fun little buglets to contend with. Some of them have been fixed, but here are a few: I'm pretty sure the comments on users.pl have the proper score now, and I'm pretty sure there are no more Score:6 comments. The reason for this is that we finally broke the Karma Bonus out into a user preference. For nearly 4 years now, the karma bonus was essentially hard coded to the users comment. But now it's actually a user preference. So if you're sick of karma whoring Score:2 default posters, you can disable the karma bonus. Honestly it would make more sense to set them as a Foe and and set the Foe penalty to -1, because a good number of Karma Bonus posters are fine, but if you wanna throw the baby out with the bathwater, thats cool with us.

Also worth noting is that a comment that gets 2 down mods loses the Karma bonus. The reason for this is that a comment with the bonus by default can't go lower than 0. Even if 3 people try to mod it down, the 3rd will have no real affect because of the bonus. This means that moderators use points on a Score:0 comment thinking that it can go to Score:-1, but the +1 default karma bonus prevents it. It creates a whole nasty cycle. The solution was simply to disable the Karma Bonus for any comment modded down twice. It's important that Karma Bonus users use their bonus wisely. Someday perhaps I'd like to enforce a stronger penalty for using the bonus to post a comment that ultimately drops to -1. Not a steep one, but that should definately be a consideration if the user is to continue to have access to the Bonus.

You've probably noticed a fair number of changes to users.pl. The layout of the user info pages is in serious flux. And by flux, I mean, Fucked Up. There are a few browser incompatibilities, and general layout screwups that really are making the users stuff hard to user. That will be cleaned up over the next couple of code refreshes.

We changed the moderation display on comments page. Originally we displayed them as a hard number (eg, Insightful=5, Offtopic=3....). And I had several issues with this. Mainly it was just hard to read on those comments that had many different moderation categories active, what with the word wrapping and all. So I changed it to be a percentage, and to only display the top 3. I find this display a much more simple and elegant view. Much more useful since for comments with few mods, its irrelevant, and with many mods, I really only want to know what the dominant couple of mods are. Especially since those are the ones that ultimately become the comment label and affect the score's comments.

The second half of this change is to add a second batch of labels that will show a user what other point modifiers are active on any given comment. The goal is eventually to make it clear what comments are triggering the long comment bonus, or the Karma Bonus, or, say the Friend of Friend bonus, and all according to your user preferences.

As you imagine, this data will greatly clarify the scores of comments for users, but it will also require some sacrifices in screen real estate to keep the whole thing understandable. So hang in there while we sort it out. This stuff is all a serious work in progress.

These changes have been some what controversial, and it will only get more controversial I promise. We've had a long standing feature request to revamp the scoring system. The existing system essentially is +1 for good, -1 for bad, and a handful of modifiers decided upon by the poster, and the reader. This is of course not a really scalable solution.

Dammit. The batteries in my wireless keyboard just died. It's 5 degrees out. I don't want to get new ones ;)

Where was I. Scalable solutions. Historically, you could look at a Score:3 comment and kind fudge together that 1-3 moderators liked it (depending on whatever obvious factors existed like AC/Logged in you saw). This creates a lot of limits on the system, most substantially is that it enforces some sort of maximum number of moderators the system can have. Once 5 or so moderators have had their say on a given comment, what more needs to be done?

What do I mean? Well, the point of moderation is to filter cruft, and to let users control how many comments they read, and have some sort of quality level associated with that number. In other words, that there will be a pyramid of comments, and if you only wanna read 10 comments, you can be reasonably assured that the 10 or so score:5 comments are the best. The problem is that if we put in double the moderators, we start getting more and more Score:5 comments. We learned that a few months ago when we doubled the number of mod points in the system. THere was a substantial increase in Score:5, but I don't think that necessarily meant that we had a better discussion- comments posted later in the discussion are still kinda ignored by moderators.

One group of people just says "Add Score:6!" but I don't really want to do that. Instead I want to redesign how scoring works. There are hundreds of ways that this could be done, and I'm certainly open to suggestions (my inbox is always available, and I'm on irc.slashnet.org much of the time). This is still very much in the conceptual stages, but I'd like to factor in many components into computing the score of the comment.

A few examples (these are all just vague ideas, nothing concrete, no flames!)

• Time since discussion creation: A comment posted 24 hours after a discussion was created, and moderated 36 hours after creation has something unique to it. Since most comments are posted & moderated in the first 12 hours, the fact that this comment bubbled up so late means perhaps it has a bit more value than a comment posted in the initial burst of discussion. Perhaps a scoring boost will help that. Part of this would be to encourage discussion to continue. And part of this would be to improve the value of moderation to those comments that number 1000+ in a discussion. Those big discussions really take a hit at some point.
• User factors: we have countless factors that we could use. UID? Activity levels? Historical Moderation Fairness? Mod points used? If UID 2,000 is a constant reader, historically moderates 98% fairly, and never lets his 5 mod points expire, perhaps this could be rewarded.
• Moderation Types: I think that the 5th 'Insightful' mod shouldn't mean as much as the first 'Insightful' mod. Some sort of diminished return. Also, I'd like to account for the number of mods in each category. I think that 6 insightfuls, 3 trolls, and 3 offtopics should have a net result of maybe +1, even tho the sum of the values of these moderations are 0, I think that the dominant moderation should have a stronger affect on the final score of the comment.
• Capping Moderation: At some point a comment needs to be done being moderated just to encourage moderation points to be spent elsewhere. As I mentioned above, it only takes 5 moderators to get an AC comment from 0 to 5. So what is the point to the 20th moderator throwing his 2 bits in? I'd much rather encourage a moderator to look at some other comment more deserving of attention? So I don't know if my best bet is to explicitly disallow moderation of comments after a certain number of moderations have been used, or to have each moderation value successively less. Either essentially has the same result.

We've also been slowly slapping in a few additional plums for Subscribers. One of them is to allow them to scroll back through comments. Currently we only let people see their last 24 comments (this is to prevent robots from crawling us, and to keep the db load down). We're letting subscribers have the next/prev links so they can browse more easily into history. Viewing those pages of course counts towards your subscription page count.

We've also set it so that subscribers can have double the friends in the Zoo system. This so far hasn't affected a lot of users, but there have been just enough people hitting the limits that we'd like to let them continue to befriend more people.

Future subscriber plums will be listed in the subscriber section of the FAQ. There are a couple of good ones coming that we're testing now. I think you'll like them. ANd as always, you can send ideas to me. IRC. Email. SourceForge. I'm usually fairly on top of all three.

That about wraps it up for me today. SimCity 4 is out, and I'm really hoping to have a few hours today to try to get a city with actual skyscrapers. So far I've not been lucky in this. And I still have friday's Farscape on the Tivo to enjoy. Also, that should give the complainers enough information to write derogatory discussions of the Slashdot Conspiracy in their journals explaining how the Slashdot Janitors are fascist demon children bent on destroying the world somehow or other ;)