Actually, they're far more curatorial when only one is allowed, because it forces users to choose this tag over that tag, and to try to develop a "culture" of what tag X should be.

I think MK's resistance to tags is not something to be discounted out of hand. Where we disagree is that his solution to his objections was to eliminate tags, while my solution was to enhance them. No bones about it - his way is a lot less work, and considering I do zero work around here, I don't really have a lot of standing to demand my way.

Personally, I think tags should be done the same way as hubwheels - something doesn't have to be just #writebetterdammit. It doesn't have to be just #writebetterdammit and #writing. It could be 3 votes for #writebetterdammit, 5 votes for #writing, 2 votes for #scifi, 2 votes for #Bradbury and 1 vote for #kleinbl00SucksBalls. Considering the way Hubski works is pretty relational, weighting one tag heavily could push that post closer to the top in that tag. Keep up #kleinbl00SucksBalls long enough and you might even get one of the most annoying gadflies on Hubski to go find something better to do with his time...

None of us wants to play whack-a-mole (troll?) with dicks, but is there an alternative?

I don't know if you noticed, but I brought literally all of Clan CIRCLEJERKERS over to Hubski with me when I showed up. They barely trolled at all. One of them, I believe, has Hubski's highest-voted, most-commented post of all time. My interactions with them have been beyond civil, they've been borderline intimate - the architecture of Hubski simply isn't one that rewards flamewars.

For that matter, SRS showed up when syncretic jumped on board. They proceeded to tag stuff with things that were immediately ignorable to me and the more prominent names got ignored by me post-haste. As such, I haven't had to interact with a single one of them.

I'm still willing to bet that a dedicated asshole or assholes bent on running the place could make a pretty good run at it - but frickin' SRS ran through Reddit like Ebola. Up in here they were more like a mild case of the sniffles. Hubski simply firewalls a lot better.

What does "scalability" look like to you?

Scalability, to me, means that you could dump 10x the userbase on it overnight and with no reconfiguration and minimal augmentation the architecture could keep up with the load. If you've got 1,000 users and you're popping along on a virtual cloud server, 10,000 users should mean nothing more than bumping out to nine more cloud servers. Obviously at some point something has to cover the cost and ideally, return enough profit to make it worth the trouble... but I could have quite a chat about that, too.

Scalability means that if you're anticipating 100,000 users nine months from now, you know exactly what it will cost to support those users, exactly what it will take engineering wise to make it happen, and an exactly linear path from here to there.

Reddit's problem is that every page for every user on Reddit is unique. If 500 people are looking at a page, the server has to create that page 500 times. There are no shortcuts. There are no repeats. There are no easy ways out. Which is why when Reddit went from 100,000 users to a million users it had constant down-time - you reach a point where any single server simply cannot render that many pages. Simply put, Reddit reached a level of popularity where there wasn't a box big enough to contain the pieces that needed to fit in a box.

For example, if you write something controversial and you get 500 replies that essentially say, "Suck my balls, loser," how do you respond?

You don't, though. There's no Pavlovian reward the way the downvote button works. Your choices are to either upvote or ignore.

The thing about upvotes and downvotes is that they permit a passive user to interact with content without having to contribute. The way Hubski does it (which it learned from /. if I'm not mistaken) you can either endorse something or not... and once something has been endorsed to ten points or so there's no real incentive to keep endorsing it.

Let's look at "the nightmare scenario."

Suppose Hubski has 100,000 users. Someone posts an AskHubski. 200 root responses within the first hour, and 350 secondary responses. Guaranteed - within 15 minutes the top 30 comments are gonna be all wheeled out.

Here's where things bifurcate from Reddit - on Reddit, all those top posts would be top posts until someone downvoted them and then there would be churn. Past about 20 minutes, though, the structure is pretty well calcified. On Hubski, however, the most wheeled-out posts aren't at the top - it's a special blend of newer, more commented and popular. That algorithm is pretty much the key to dealing with large comment subsets and Reddit used to fuck with it a lot.

Here's where things could easily bifurcate further: Reddit gives you "new" "top" "best" and "controversial." Hubski could easily put its parameters on sliders and throw them up on the user page... or even on the comment page itself. Let users fuck with it to get what they wanted. Monitor what people are setting it to and do a normal distribution on that. Reset the algorithm weekly (daily? hourly?) to show what's working... or shit - do it per post.

I like pushing The Starfish and the Spider on people to demonstrate what can be done when nobody's in charge. Reddit should be like that, but it isn't - it's a top-down hierarchy run not dissimilarly from North Korea. Yeah, the moderators try to keep things functional like good citizens, but they're completely powerless in the overall structure. They're just cannon fodder.

If Hubski tries to figure out a way to make large comment sets parseable without having to have a human being beat on it, it will win the world.

Period.