The way D2 was rolled out for testing was that we actually had two discussion systems in place for awhile. One was a University of Michigan Research project written by nate. We used that as a rough framework to build our system. Subscribers got our system. Odd numbered users got our system, even got ours. Nate's system was given to a few hundred users, while ours was given to an ever growing number of readers that we rolled out over the course of a couple months. This is how we roll out many functions that aren't ready for prime time. Giving everythign to everyone all at once has performance issues for us. A slower rollout makes sure that we can work the kinks out before we give it ultimately to anonymous users.
If you use the handheld/low bandwidth options, the floaty control widget can cause problems. We know. It's on the list, but since only a tiny percentage of users are using those modes, it's not a top priority. The fix is relatively simple: the floaty needs a toggle from top to side. The code has a widget for a top floaty, but it's an older version. Basically I want to make the side floaty work properly (drag & drop instead of clickable for example) then redesign it to work up top and add the toggle. But that problem is easy to solve.
A VERY key point that a few users got is that this doesn't solve the problem that the old system has in favoring older threads. There is no good solution to this, short of randomizing top level comments. The REAL solution is to rework the scoring system to re-value threads with more granularity then -1..5. That of course is the plan.
As for one-click moderation, this is a baby step for us. The new moderation system is vastly different then the old one... and in-place moderation was critical to make it work.
A number of readers commented on the highest comments first not existing in this system... thats true, but you can fake it reasonably well. By setting the threshold fairly high, and hiding a good number of comments you can easily filter to score:4 or score:5 comments. Admittedly that doesn't give you 5s then 4s then 3s, but for mature discussions there are already dozens of comments. The new moderation will aim to address this problem in more detail, but I think a high enough threshold is BETTER then simply sorting by score becaause it's possible for you to navigate up or down without a page load when you do find a comment that you think might be worthy of further research.
As for patches, tf23 and a few others noted that we don't provide much in the way of direction. Well thats kinda true- the SF project page has tons of feature requests. If you emailed me and asked for my opinions on any of those features i'd tell you. I simply reject all feature requests that are worthless. Some of them are low priority, or things that I just wouldn't use for Slashdot. I read the mailing list- although I have little to say there. I've learned that if I'm not paying people, I don't usually get what I want for Slashdot, and thats fine. Code what you think works for your stuff. But if you want it on Slashdot, ask me how it should work
Anyway the feedback on D2 is mostly favorable. I'm disappointed that the bulk of the discussion yesterday focused on the fact that we've chosen to beta test on the single browser used by most of our readers and worry about compatibility later. It meant I got less real feedback then I might like. But I think that's always the case- if there is one glaring issue, readers can't see past it to talk about the hundred other issues that are honestly more helpful...
What's interesting to me is figuring out how users can navigate this datastructure intelligently. What does it MEAN to expand a thread. To collapse a thread? Do I need siblings? Do I need expand-all? I don't want 35 buttons... where is the line between needless clutter and the necessary UI? Interesting problems to solve. Fun stuff. What's tough is that some readers want everything and the kitchen sink, but most users will find 3-4 buttons sufficient (and others will think 2 is intimidating). We have to balance minimalism, functionality, and hardware limitations. I find it enjoyable.