Hacker News Reader

Comment by sologoub | Hacker News Reader

Although I get the concern here, it would seem that if you removed the nofollows, it would open up the site to all kinds of abuse and diminish the value of the content very quickly. It doesn't strike me as a very difficult task to create a few thousand profiles and have them vote for each other (and also give you a ton of votes to random people to hide better), and then use the collected power for SEO spam.

I agree with you that there has to be a better solution, but it doesn't strike me as a very trivial one... any thoughts on how one would approach this?

View on HN

>It doesn't strike me as a very difficult task to create a few thousand profiles and have them vote for each other (and also give you a ton of votes to random people to hide better), and then use the collected power for SEO spam.

I have to disagree. I bet this kind of abuse is well studied by any major site relying on a reputation system/user content. I can't point to anything specific, but off the top of my head discovering that kind of abuse seems exactly like finding strongly connected components: http://en.wikipedia.org/wiki/Strongly_connected_component

So you could probably prevent this algorithmically, but even if you couldn't... the site is heavily moderated. Bad questions and answers are down voted or closed. In my experience, getting lots of reputation on SO actually requires a lot of persistent effort. I have just over 1000 rep myself and it felt like it took forever to get there. Even if you could get lots of upvotes, you'd still be looking at daily reputation limits. And finally, you'd waste all that on one or two spam links only to get perma-banned and have all your links deleted?

A combination of high reputation + an initial nofollow time period (to allow spam links to be discovered) seems like it would be pretty effective.

There's really only one route to taking advantage of the google juice: http://xkcd.com/810/

While I agree that it would be possible to get these networks exposed, it is still non-trivial effort. That's really the point I am making. Personally, I'd rather see them allocate more resources towards making developer QA even better, than chasing protective features.

My 1781 reputation was not that hard to come by - it happened organically with me trying to help out in the Google App Engine section. I'm not a developer by trade, just a hobbyist.

Based on that experience, I'd imagine there to be areas that are much more obscure, with little moderation oversight, being decent sources of easy reputation points...

"it would seem that if you removed the nofollows, it would open up the site to all kinds of abuse"

That's why the author of the article suggested that they remove "nofollow" for users who have an established reputation, like SlashDot does. That algorithm would seem to be very easy to implement.

What happens when a less established user edits the post of a more established one ( and maybe also vice versa)

You need a good amount of karma to edit someone else's posts, so that's a non-issue. If it was an issue, the rel=nofollow should be based on the rep of the user creating the link, whether that was the poster or the editor.

You need 2000 rep to do this, which is not so easy to gain, but certainly achievable for a highly determined person. However, what @wookietrader wrote applies -- each edit is marked with the person who has done it which solves the problem.

Anyone can edit. If your rep is low, it has to be approved.

Since you can check what user added what link (using diff), you can still use the reputation threshold. If that does not scale, just use the minimum reputation of the editors' reputations.

Interestingly they already remove nofollow from links in your profile after you reach a certain karma threshold

As the article mentions; at a certain reputation level StackExchange users are granted a followed link in their profile. Similar metrics could be applied to questions and answers with the "nofollow" being removed from highly rated user content which is unlikely to be spam.

Still a hard problem but author reputation and question/answer votes give a good set of data from which we might identify links to sources which deserve attribution.

The article seems to argue that because there is no perfect solution to spam (i.e. spammers will still want their links on SO even if they are nofollowed), then there is no point in marginally reducing the incentive to spam.

I think the point is myopic. Human moderators will eventually clean up spam, but Google is freakishly fast at indexing things like StackOverflow. Omitting rel=nofollow would mean that SO ends up lending linkjuice to spammers at least some of the time, thereby providing spammers with incentive to relentless spam even if their spam is cleaned up by humans.

My general rule of thumb is "any content that comes from untrusted users gets a nofollow stamp". It's "XSS prevention" for your pagerank.

But SO knows which users are trusted, and which aren't. So why is it adding rel=nofollow to trusted users?

When I say "trusted users", I mean "people who have the product's interests before their own". I don't think this applies to users, even high-reptuation users.

Arguably, it eliminates incentive to link to anything except the most useful resource. I have >11k reputation on SO, and I'll easily admit that I'd be tempted to link to my blog or whatnot for an answer whenever I could wedge it in, if I could siphon off some of SO's linkjuice. I'm not interested in spamming or anything, but that's how the game is played; SO has such a massive amount of clout with Google that any non-nofollow links from it confer an awful lot of weight.

By just nofollowing everything, they disincentivize people from even being tempted to game the system and promote their own stuff, even if it were done in a helpful way. Instead, I am incentivized to provide the most useful, comprehensive link I can find to answer a question. All my incentive is in earning an acceptance of my answer, rather than just in having an answer good enough to not not get spam-binned that sits there and feeds me pagerank.

I have seen reputable open source software developers put paid linkspam on their personal websites. When I called them out about it, they said "ha ha, I was wondering if anyone would notice that, yeah, they paid me a pretty good amount of money for that."

Trusted contributors could easily be bought off. Someone who was sick and tired of StackOverflow after putting way too much time into it, who was offered the right amount of money? Hey, you know, it might be worth it to make a few bucks instead of a few more meaningless internet points.

If you just make the threshold so high that only the very top contributors are nofollowed, then only that small cabal controls where SO's considerable linkjuice goes. If you broaden your criteria, there are more and more potential corruptible people; and it becomes easier to generate a little quick and easy reputation with a sockpuppet that you can spam with a few times before you're caught and move on to another sockpuppet.

I'm confident that stackoverflow already utilizes voting ring detection logic. High karma on SO is quite valuable, I'm sure there are many that have already tried to game the system, people who are probably a lot better at evading the voting ring detection than the average spammer...

I'm curious how karma on SO is valuable. Not in the "you seem smart i'll give you a job" sense of value, but in the actual converted into dollars value that spammers go after.

It's valuable in the "I'm a leet hacker" sense.

Reply on news.ycombinator.com