There was a cool link curator service called StumbleUpon. You click a button, and it takes you to a random interesting web page. One of my favorite things about this is you can comment on these pages as well. It was fun to see other peoples reactions to random pages, flash videos, memes, art, essay, etc. Eventually I realized that you can click the 'discussion' button even on pages that StumbleUpon didn't bring me to, and there was still a discussion much of the time(The link must already be in the database, of course).

Anyways, I've always wanted to find something like this again. This sounds similar. I'm gonna try this out and take a look. The big thing that makes it work, IMO, is having enough people participating and having comments be moderated. Unique, sincere comments make the experience fun. Spammy, trolling comments make it just every other boring corner of the internet.

I loved this, I could go anywhere on the web and have a comment section. The important thing about this, though, was that the comments were usually pretty relevant. Rarely did I see trolling or off topic comments(Ie politics). It'd usually just be maybe a one line impression of the site. It was perfect.

> One of my favorite things about this is you can comment on these pages as well.[...] Anyways, I've always wanted to find something like this again.

https://web.hypothes.is

(The public stream is here: <https://hypothes.is/search>. Also supports parameterized RSS <https://web.hypothes.is/help/atom-rss-feeds-for-annotations/> and has a developer API—info available on the Developer tab in your account settings.)

I too remember StumbleUpon fondly. Now I feel like it's a way to get people to go to questionable sites, but back then it was a lot of fun.

> Rarely did I see trolling or off topic comments(Ie politics).

I wonder if this is due to the format or strictly because of the time?

It wasn't completely randomized though if I recall, people used to submit sites to it and would tag it in different categories and then when you signed up you would select your interests and then it would choose things from the possible options.

I think it was partly the time and partly the community. The format probably didn’t have much to do with it.

Yeah immediately thought of StumbleUpon.

Incidentally, this is one of several things I've thought FF should have integrated into the browser to differentiate themselves and maybe retain or, if you can imagine it, grow their userbase. It'd be an outstanding core of a distributed social network, if they wanted to take it that direction, and a decade or so ago FF was uniquely positioned to make such a thing happen by integrating it directly with the browser so the best way to experience it was in FF. These days, I'm not sure they've got enough users left to make such a play, but they did at one time.

Yeah, cold start and moderation are two problems I need to solve.

I'm focusing on documentation pages at first hoping to build a commpunity that is helpful. Like pointing out pitfalls and better solutions to existing problems.

Also I used to be a stumbleupon user too. I don't remember the discuss feature, but I remember using it to find hidden gems of the web.

I'm not an expert of any kind, but I'm going to give my unasked-for advice anyways. 1. Paid moderation as the main solution probably isn't feasible. Even if you can find a way to monetize your project without annihilating your userbase, the amount of text will outpace the manpower you can afford. 2. Moderating is a crap job and there are never enough volunteers. Because there are never enough volunteers, you will feel forced to accept subpar volunteers and tolerate ones with annoying proclivities. The solution to these problems is the same - you need to make the manpower you have go further (cover more text). First, minimize the need for moderation. That's a difficult task when your entire project is user-submitted text, but if you can prevent people from creating multiple accounts and ban or suspend people who generate disproportionate amounts of moderation, it'll help. Second, you need good moderation tools. Don't be reddit, where moderators rely on third-party tools and bots to manage their community. You need to provide good tools to detect things that need moderation and not bother humans for things that don't. The more you can force-multiply your paid and volunteer moderators, the better you'll do.

p.s. I could ramble about moderation for a while. I'm in favor of suspending rather than banning users in most cases, and I've been toying with an idea that suspensions could be based on the amount of content and crap someone generates. You want content, and you want to keep the level of crap in that content low enough your moderation can keep up with it. If someone generates a lot of content and a little crap, you can just remove the crap and only give them a small suspension. If the proportion of crap to content someone generates is over whatever threshold you think is manageable for your platform, give them increasing-duration suspensions so they're generating less content and less crap - keeping your sitewide crap threshold down. If their crap-to-content ratio goes back below the threshold, or your sitewide crap load has gone down, you can start cutting down the duration on the suspensions. An algorithm could surely be devised for this and then you just set a 'crap threshold' platform wide and moderators don't have to do anything other than remove a post and the suspension is automatically applied. There are exceptions, of course. Illegal content? Removal and warning if it's not too serious - ban if they do it a second time. Removal and ban if it's serious. Harassing other users? Ban. Crap of a type that will drive away your userbase? Warning, ban. You could probably devise a more strict algorithm for this, but when it comes to people who make a community worse by being there, being lenient with them is something you're doing for their benefit, not yours. Whether you choose to be lenient is up to you, but if you get it wrong they'll drive away the people who generate content and you'll be left with people who generate crap.

Thanks for the comment. It was quite insightful.

I don't have any prior moderation experience, so I plan on taking up the task myself initially to get a good grasp of the challenge. That way I can build good tools to make the task easier.

This webpage has the story of what happened to StumbleUpon, a link to it's successor Mix, and some alternatives:

https://techboomers.com/websites-like-stumbleupon

I've dreamed of something like this for a long time.

However, the actual tool is only a tiny part of the problem. The bigger part is moderation and spam.

While this is niche that's not a big deal, but it's also nothing like reddit in that case.

What if it does take of? How will you handle spam/astroturfing/hate/illegal comments?

All of that is valid, but you shouldn't worry about it until it actually becomes a problem.

99% of the struggle for any social network is getting people to use it at all. Those problems you list would actually be great to have to deal with, because it means the network is taking off.

Lots of problems to be solved yet. Moderating 10 comments / day is very different from moderating 10k/day which in turn is different from millions a day.

I'll initially start by moderating comments myself, (need to create a policy first) and see what to do from there.

Moderation should work peer-to-peer. If somebody wants to post they have to first review posts which were flagged by users to violate the TOS. If sufficient users agree on the flag, the post is taken down.

This. You should read up about how recaptcha worked in the early days to do 'authority of the masses'. (I don't know any particular sources, maybe these methods are properly defined now in some framework.)

I would absolutely love to see this attempted.

In order for a user for flag a post. They must highlight the portion of the ToS that is violated.

This has the added benefit of requiring people to actually read the frickin ToS to use the app!!!!!

Quick note-- https://gist.github.com/ritinkar8/4991a0770aa0363598aa7c5795... landing page I'm taken to after installing looks incomplete. Just a FYI in case you didn't know.

Yeah, I need better onboarding. Will fix.

How do you handle new releases of the documenation? A bunch of now irrelevant comments, or are they all summarily deleted (or at least restricted to the previous version)?

So this is not a problem I've solved yet. But I think the solution would involve a mix of community voting and good sorting algorithms.

But I'll only know the correct solution once I start tackling it.

Hope this works out!

I used to read the comments on PHP.net to learn more about certain features or to find example usages. This made the PHP community seem very approachable and helpful.

Similarly! You always wanted to read the comments, because the docs would not necessarily tell you that such-and-such a function had tricky edge cases, or had been replaced by a better version, or was a massive security risk.

Of course all of that should have been in the docs to start with, but the comments really saved that site.

And if you're looking for a quick walkthrough, it's at: https://www.youtube.com/watch?v=oln-KdIczOM

lovely! aaah i have such a soft spot in my heart for ideas like this. reminds me of "hoodwink'd," a similar project by _why the lucky stiff from 2005 that let readers leave comments on webpages tagged by URL. (see http://web.archive.org/web/20080106065546/http://hoodwinkd.h... )

Back in 2010, I also ran my own project in this space called "Goggles." https://knowyourmeme.com/memes/sites/goggles It was a bookmarklet that turned any webpage into multiplayer MS Paint; viewers with the bookmarklet active could see others painting on the webpage in real-time, before websockets made this kind of interaction commonplace. The entire thing ran on a single-threaded nodejs instance, serving something like 1k simultaneous connections at peak. Every half-hour or so it would crash from memory leaks. Good times.

Moderation was the thing that killed my project. Some users became victims of targeted harassment campaigns; bullies would draw dicks, swastikas, nooses, racial epithets etc on their personal tumblrs or homepages. There's something insidious about knowing that folks could be graffiti-ing your own webpage behind your back unless you're constantly checking goggles, so it didn't feel right to keep the project running knowing I didn't have the motivation to add proper moderation tools and didn't want to build user accounts or any sort of reputation system to keep the UX simple.

Some folks made absolutely stunning pieces of art on corners of the web that nobody saw. I probably still have the shapefile database archived somewhere... maybe we could reunite artists with their work someday.

This is very neat. One question, would you consider making a version of your service that works like discourse.org? I don’t mean changing the discussion software to match theirs, I mean allowing clients to embed your software in their page. Also, for what it’s worth, if you had plug-in authentication of users you might find some clients in in web3.

I haven't thought about that direction yet, although I am aware of discourse and disqus.

Metadocs is pretty young, so I'll think about it.

I've dreamed of something like this for a long time.

However, the actual tool is only a tiny part of the problem. The bigger part is moderation and spam.

While this is niche with a tiny user base that's not a big deal.

But, what if it does take of? How will you handle spam/astroturfing/hate/illegal comments?

Totally tangential, but HN should be able to handle duped comments.

Genius was working on something similar for a while, but I don't think it ever took off: https://genius.com/web-annotator

Cool!

How does it identify elements to anchor to, and what happens when the original text changes?

If you're asking how the "Highlights" feature works, its using something called text fragments.

When the original text changes then the highlight is dead. Meaning you can still discuss it, but it will no longer be highlighted in the window.

You can read more about text fragments here - https://web.dev/text-fragments/

That seems like it would be a problem for something like a documentation site, where even small changes to the text would break the entire highlight. Maybe you can implement something via diff to maintain a fuzzier anchor.

Didn't gab do this exact thing years ago and it caused a major uproar?

This reminds me of the old shadow run rpg books. They had in character comments throughout the books. There were whole subplots and narrative arcs just in the comments. It was great.

Any plans for Firefox (and does this need to be an extension at all - there are some rather powerful website extensions that load via JS scripts)?

Yup. The code was written with portability in mind. As soon as I get time to test it on Firefox I'll do it.

And have you thought about whether or not it could be loaded with javascript?

That could eliminate the need for platform specific extensions.

There's a service worker in the background that does quite a bit of the heavy lifting. So, it can't be a pure JS script.

This looks very useful. Do you need a user account to use it? Where are the comments stored?

You only need a user account if you want to post a comment or vote on one.

You can browse as guest.

The comments are stored in metadocs servers.

This is pretty cool, nice job!

There was something like this going around in ~2008, it was a plugin bar that turned the internet into a game, you could leave mines and portals on websites that would take you to other websites, so people could add their own hyperlink layer on top of the internet.

I wish I could remember the name of it, I played with it for a few weeks but then they put up a paywall.

That spirit of communal being-on-the-internet I only saw breifly recaptured again by the Beaker Browser, which unfortuantely seems also to be going the way of all things. There you could fork and remix any website you were on, could in principle afford a much more mucking-in stle of what you are trying to do here.

I think this could be great! Keep at it!

this looks super cool and I would love to try it! any hope for Firefox one day?

Yup, there most definitely will be. Hoping to do it within the next few months.

which problem are you trying to solve?

It's hard to articulate the problem like that.

So I was going through a documentation of an ORM library for another project of mine, and it was kinda poorly written. And I didn't feel like going through the effort required to go to another website and ask questions there. I thought I wish I could ask questions right here.

So I took that feeling and turned it into something tangible.