Show HN: Words for Chrome – Smarter, safer web comments | Hacker News Reader

Show HN: Words for Chrome – Smarter, safer web comments

words4chrome.com

23 pointsfivedogit55 comments

Cool, I've thought about something like this for ages, but lazy always prevails. The only thing I don't really agree with is:

"Let’s face it. One major problem with embedded web comments is that everyone is invited to participate. They’re too open.

The WORDS community, on the other hand, is necessarily comprised of people who (a) use Chrome and (b) desire a better commenting experience. Why else would they have installed WORDS?"

That's true for now, but any system that gains any amount of steam will inevitably pick up trolls, flamers, and generally idiotic people. You can't stop it from happening. And I don't know that anyone has ever found a good solution for it. In fact, the whole 'Top Comments' and up/down-voting thing most third-party comment widgets employ is literally to combat that problem, in the hopes that garbage falls to the bottom. It obviously doesn't work perfectly and I do still agree on the whole with your assertion of it being overall detrimental to good discussion.

Anyway, I'm guess I'm mostly curious if you have plans for the future of if/when the extension gets more popular and you do start finding discussions bogged down by trolls/spammers/etc and what ways you'd try and combat such things when you have 100k+ users or whatever.

I've thought about something like this for ages also. But lazy always prevails.

Now, do you know something about a Google product from some years ago that did something like this, but was discontinued?

Here's my explanation of Google SideWiki... near the bottom.

http://www.words4chrome.com/eli5/

Q: “What happened to it?” A: “It had several problems. (1) sidebars are the worst (2) web site owners hated it, presenting a threat to Google’s core business, and (3) because Chrome had negligible market share in 2009, it had to function in all browsers at once, harming UX. Hence, the sidebar. And the name."

Also, it was open to everyone. Words is designed to select for -- um -- not everyone. Internet explorer users, for instance.

Precisely. I've been trying to remember that name for months.

Did you mean Google Wave? It was open-sourced as Apache Wave.

I don't recall Google Wave having website commenting features.

It was a federated protocol for real-time editing, one could layer various workflows (group editing, email, IM, wikis, comments, etc) on top of that protocol.

Thanks for the thoughtful feedback.

One of the problems with other commenting systems is what I call The Clean Slate Effect. If a troll is sucky on site A, then goes to site B, they start over with a clean slate and can continue being sucky.

Words benefits from the fact that it's web-Wide. If a user is terrible, they get silenced everywhere.

You're totally right (as was Paul Graham when he spoke about how Hacker News is evolving) that these sorts of systems get worse the larger they get. And that getting too big too fast is definitely a bad thing. Just look at Digg.

So I definitely anticipate badness down the road, but feel the code foundation is there to deal with it as it comes.

>One of the problems with other commenting systems is what I call The Clean Slate Effect. If a troll is sucky on site A, then goes to site B, they start over with a clean slate and can continue being sucky.

How is this different than any other 3rd party commenting system like Disqus or Facebook?

So. Many. Reasons. Just give it a try and you'll see.

> these sorts of systems get worse the larger they get. And that getting too big too fast is definitely a bad thing

I was thinking about it - why not add a view mode displaying only the comments of the early adopters? On reddit, for example, it would be interesting to see only comments of users with accounts 5+ years old.

Gawker's network effectively does that by focusing on only Featured Comments when you scroll down and making you click through to the plebs.

I thought about this idea -- only displaying comments from users with high ratings or whatever -- but decided it would break conversations. Users would see comments, but not some responses or worse, some responses without the original comment, etc.

> necessarily comprised of people who (a) use Chrome and (b) desire a better commenting experience

That also means that it will be a ghost town, and probably never bootstrap to a decent user base with interesting discussions.

http://viewsourcecode.org/why/redhanded/images/hoodblink.gif

I miss _why.

for those who don't get this: hoodwink'd[0][1] was this experiment/hack/art project by _why some years ago, which allowed implementing a parallel unified commenting system for the web.

It was super cool because you'd go to, say, hackernews or slashdot and if you were part of the network you'd get access to a different set of comments, but integrated in the website rather than as an external popup (individual sets of stylesheets+js had to be written for each website of course).

For example, here you'd get the blinking hood gif, and on click you'd see new comment threads appearing with red usernames. Integration came through, iirc, a custom proxy server or a greasemonkey script.

It was super cool.

[0] https://github.com/whymirror/hoodwinkd [1] http://ecmanaut.blogspot.hu/2006/01/hoodwinkd.html

Nice explanation. (:

Consider removing the downvotes? Keep only the upvotes?

Downvoting IMO feels bad for the poster. They stop feeling like participating. They might have very good and insightful posts which might not be popular. The downvotes end up driving them away and all you're left with is an echo chamber of groupthink

As someone who's received a nontrivial amount of downvotes in this community and others, I can confidently say that I appreciate them as a mechanism. Almost every downvote that I've received was for a legitimate reason - a few were just out of disagreement, but that's the exception rather than the rule.

In a way, it does create an echo chamber: when the community consistently downvotes things that don't conform to X (in HN's case, X usually being "well-reasoned commentary relevant to the thread"), people tend to shy away from those things. One could consider the utter lack of response gifs on HN "groupthink", in the sense that everyone knows the community is unfriendly to them so they don't post them (even when they're funny).

A well-organized hate group can use downvotes to silence reasonable discourse and trumpet their ideology, so it looks like the popular opinion when it's nothing of the sort.

You see this happen on Reddit. A hate group called SRS does it. They've even been leaking onto Hacker News recently, with some of the same lines they regularly use.

Wow, that's really disheartening to hear. I've personally never seen that around here (thankfully), but now that I'm over my minor obsession with Dogecoin I don't spend much time on Reddit.

For my money, downvoting is entirely necessary. You know how a comment like "R u srs? Luv this vid!" is normal on Youtube? I want users to downvote the living hell out of that comment. I want the WORDS community to hold each other to a much higher standard.

Downvotes seem like the only way to actually maintain order.

There's nothing disorderly or actively harmful about that comment. It just adds little value compared to the search space of all possible comments. So it seems semantically more sound to simply withhold upvotes from it, rather than punishing it with downvotes.

Let's say a comment on one site gets 100 upvotes and no downvotes. On another site, another comment gets 10 upvotes and no downvotes. Was the first more popular, or just more trafficked?

You need the ratio of up-to-down to determine a comment's true usefulness.

You can't compare upvote numbers (or even upvote/downvote ratios) meaningfully between sites, because their communities aren't the same. Your proposed scenario tells us nothing beyond pure numbers -- 100 people voted on one, 10 on the other. You can't say anything comparative. You certainly can't speak to strength of intent.

This is very sketchy. When I try to login with Facebook, it shows me chromless modal window with (what appears to be) a facebook login page. Except that I'm already currently logged in to facebook so this shouldn't be necessary. Sorry, I'm not going to hand you that info.

The Google login window seems to be safe, but I'm intrigued about the how and the why.

Historically, Chrome has not handled third party login very well, and one could argue that it still doesn't.

However, they recently took the chrome.identity package out of experimental mode and it's what I'm using for Oauth flow because it handles authentication in the browser, not in cookies and URLs.

I agree it looks sketchy and several of us have complained to Google to make it look a bit more legit. But you know how that goes.

If you're really concerned, crack open WireShark and take a look at the handshake yourself.

Is there way to register without Google or Facebook account?

No. See my dissertation about it here:

http://www.words4chrome.com/2014/about-words-login-options-a...

That doesn't explain why it doesn't have a native log-in, though ( even though the preamble says it will ), or support a general protocol such as OpenID.

Here you go.

http://www.words4chrome.com/design-choices/#nonative

This is my 4-year, off-and-on passion project. All suggestions and feedback are appreciated. I'd especially welcome product/feature-based ideas about how to overcome the enormous cold-start hurdle.

Note: Re-posted with permission from Dan. Last week's post got dinged by a semi-false-positive algorithmic penalty after only a few minutes.

Would you consider supporting OpenAnnotation or similar efforts from W3C?

Cold-start: convince a site with an existing audience to promote it.

First I've heard of OpenAnnotation. I'll look into it. Thanks.

As far as overcoming the cold start, I'm looking not so much for marketing ideas, but for product or feature enhancements that would be valuable to users immediately, without a huge number of other users.

Google search - useful for 1 user Facebook - theoretically useful for a handful of friends Twitter - useful for a slightly larger network

Words doesn't have friend/follower mechanisms -- it's just not the nature of the product. The intention is for users to interact with each other as they happen upon the same content. It's a dynamically-changing social graph. Unfortunately, this means the critical mass of users is definitely north of 100k, probably higher.

It wasn't a marketing suggestion. If an existing community (even a small one with hundreds of users) promotes the plugin, you will get feedback about requirements and feature enhancements. This is a necessary precursor to finding 100K users. Even Facebook started in a limited school environment.

These "second screen" or "third voice" apps require people to be (a) trained on the tool, (b) have a common set of target objects that they want to talk about. You could start off simple by looking for a small community of people who:

(a) use chrome

(b) regularly visit multiple (5-10) sites related to their primary site

(d) see value in having their existing conversations appear in-context on the secondary sites

The plugin may have business applications for proof-reading.

Ok, so if I understand this correctly, in order to be certain I'm not tracked, I need to host my own back-end server? And the server stores the comments? And if I'm hosting my own server, I'm likely to end up talking mostly with those that sign up with "my" version of this? And if I want it to work in eg: Firefox I'd have to port it? Finally, wrt. tracking -- if the comments are stored on the server (and that's a REST service or something like it) -- wouldn't it be trivial to log activity at the routing level (eg: ha-proxy/nginx/other front-facing web server)?

This isn't meant as negative criticism, just trying to figure out if I've understood how it works correctly? (Note, I appreciate that the source is available, but obviously I can't know that the source you share is the source you run on your back-end. So we're back to trust, which is fine in my book).

[edit: re other clients -- the most straightforward thing to do would be to make an iframe(or js)-based proxy service (a la stumble upon) in order to enable any browser, I suppose?

I've thought about something similar, and I wonder if an XMPP back-end with one "room" pr url and the server-side log-extension might work? That should then make it trivial (unnecessary!) to enable better clients than web-only -- ie use a command line xmpp client to participate in the discussion, rather than forcing the use of the web browser. I realise many people think using the web browser is an advantage -- I generally find GUIs built on top of browsers to be slow and not very efficient. Extensions less so -- but the problem is of course that you'd have to port to Firefox, Opera, Webkit and IE (along with mobile variants) -- to achieve parity with systems like disqus etc]

The problem with XMPP rooms is that if you can connect to them with a chat client, people start using them like a chat client. Which is not necessarily what you want.

A valid point, but it's generally what I want :-) Well, at least as long as I can create the right kind of echo-chamber of like-minded people... ;-)

Perhaps XMPP+Server Log extension+rate limiting?

The inability to edit messages might be a worse strike against using (a not too heavily modified) XMPP server, though.

You're exactly right. The backend source code is here: https://github.com/fivedogit/words-backend but there's no way to know for sure that it's the same code running on the backend other than my word.

Seems like this might be a good business for someone to start, matching a github repository to a running server; third party verification that said code is actually what's running on a server.

If it helps, I made an "about me" page, so I'm not quite a total stranger: http://www.words4chrome.com/team/

This reminds me of a product from the 90's where it let you leave post-it style notes on any website. You could then turn the nodes on/off for any given site.

Anyway, that was a great product, so I'm glad to see that you're thinking around similar lines. A meta-commenting system that runs any site is a great idea.

The Google login is a bit too disconcerting for my liking. At the very least showing the url bar would've have been less unsettling.

Only by opening the chrome inspector was I able to verify that it was indeed a valid login page and not a phising attempt as it appears to be at a cursory glance.

@fiveogit: Love the idea. Need to be prettified though. How would I see comments on mobile devices or non chrome browsers? Can I inject with JS into my website? That was some of my first thoughts :)

Exactly my thoughts. I use Chrome at home while I can't get around Firefox at my workplace - is there any way to at least fetch the comments by different means?

At one point during development, I had the extension actively replacing the Facebook comments on TechCrunch like so for visual stimulation purposes:

http://www.words4chrome.com/wp-content/uploads/2014/05/tc_zi...

Could I create a box that a site owner could embed into their page. Absolutely. But why would a site owner do that? WORDS is built for users, not site owners and would lack top-down moderation controls, etc that LiveFyre and Disqus provide.

That's why I've held back on building an injectable widget. it's dissonant with what WORDS is supposed to be.

From a skim of the source, this extension appears to ship every URL you visit to a third party. (It does so to fetch the comment count on the current URL.)

The backend code is open source and available here:

https://github.com/fivedogit/words-backend

The URL is required to retrieve the proper comment thread for that page, but is never associated with the current user.

Look for the line "else if (method.equals("getThread"))" in co.ords.w.Endpoint.java and you can follow the url variable from there.

I like the idea though I'm on opera.

Edit: oh and it requires google login. No thanks.

Here is my answer to the Google/Facebook login question:

http://www.words4chrome.com/2014/about-words-login-options-a...

and

http://www.words4chrome.com/design-choices/#nonative

Even though there is no privacy downside, people are still afraid of Google and Facebook logins. I guess there's no way to change that perception.

Just out of curiosity, you and several others have singled out Google, not Facebook, for criticism re: the login. Why?

Good luck getting traction. The web is littered with the corpses of "comment on any website" apps. The problem is that there are too many websites and too many incompatible "comment on any website" standards for any one of them to emerge as a clear leader.

P.S. You list non-tracking as a differentiating feature, but how do you silence speech you don't like without tracking users?

W3C is tackling the compatibility issue.

http://www.w3.org/2014/04/annotation/

https://hypothes.is/blog/peer-review-annotation/

Listen to Walter, you don't want to be an island of comments in this case. Following annotation standards will allow this content to be gathered and reused elsewhere, for things like fact extraction, sentiment analysis, meta-discussion, +++, it will also allow others to immortalise your content if you happen to fail.

On the other hand, maybe that's exactly what you don't want. My personal pet peeve is comments on scientific papers, where openness of this content is not only desired, it's crucial. Your situation might be different.

'The web is littered with the corpses of "comment on any website" apps.'

There is one "comment on any website" tool that has existed from the beginnings of the web and still thrives. The way it works is, you put the comments on your own website.

Do you mean, post comments you have about another website on your own site? Definitely works, although obviously harder for others to discover your comments compared to a single commenting tool which everyone used. Although I'm sceptical such a tool will ever exist.

Everyone wants to be Google Sidewiki

Reply on news.ycombinator.com