Comment by gwern | Hacker News Reader

Comment by gwern | Hacker News Reader

Parent Full thread

It actually is broken in a lot of ways, and I don't mean the complaints about stalkers or fake reviews or other community problems, just on the pure technical level. It's been slowly bitrotting, it feels like it somehow gets slower every year, and they've been removing features. I recently moved all of my stuff off GR and stopped using it because I asked myself why I was putting up with it when it clearly was only going to get worse over time.

For example: lists in reviews don't render, somehow they broke list markers, and this has been the case for like a decade now (?!); you can't add links to profiles anymore, and you can't edit your profile if it already has a link (because 'spam'); you can write book reviews which you can't then edit (because the edited, but somehow not the original, violates 'length limits' - which are shockingly easy to run into if you include any links); they disabled part of the export API recently, and I wouldn't be surprised if in a few years you can't even export your books...

Gets slower? It loads every page I tried in probably under 2 seconds? Is that slow these days?

Yes, it is. The popularly cited studies these days suggest an upper limit of one second to ensure a user maintains flow.

And consistent <250ms will feel like magic — try pinboard as an example of lightning fast UI

Pinboard doesn't have near the traffic or complexity as goodreads. It's basically fetching text out of a database with a user-id lookup. No multiple databases for books, user profile, reviews, etc.

I want to assume this is tongue-in-cheek.. but I'm not actually sure?

The whole page. The text is available almost immediately, so not tongue-in-cheek. Who the fuck leaves a page that they went to if the see text and a page don't see all the pictures in 2 seconds? We have a shorter attention span but not that short.

I meant your comment. I think I have my answer.

So a new goodreads minus the technical problems comes along.

It gets popular. The community becomes a mess, it becomes costly to police, you have people upload an ID and the world hates you for invading their privacy, management shifts resources away since its stable-y making money regardless (given some Byzantine model that matters to ownership anyway) and then it falls victim to bit rot.

So a replacement to your goodreads replacement comes along...

You can do ID (or just validate a few other ways) with anonymous accounts, done properly seems a good balance.

Some would hate it, but if it actually improves the quality and the company has some security chops I’d see it as a selling point.

By definition, you cannot have ID for anonymous accounts. Not without a third party (who I want nothing to do with anyway).

Furthermore, is it really up to a book review site to solve the problem of identity and anonymity on the Internet? Seems like the wrong place in the stack to focus on that.

Yea, I think it's a good model for other type of companies and have thought of it before which is why I brought it up here, just felt like clarifying it is possible to do if desired.

You can actually do validation for completely anonymous accounts. The most common version is DDOS protection where even read only websites can still benefit.

An anonymous review website could similarly rate limit how quickly reviews change, so someone spamming 1,000 reviews accomplishes little.

Actually, a decent anonymous single login you could do today roughly is basically TouchID. You’d have to implement it yourself with ML, a webcam and some client side code that hashes it.

If it worked well, I’d use it in a snap.

Communication issue:

I'm defining "anonymous to the world" or "anonymous publicly", whereas you're defining it as "anonymous to everyone, even the company".

But in the scope of a book review website, and this thread about preventing spam by having ID enforced, my comment made sense as that. The company knows you, but you can have an anonymous handle to the world. I had clarified that with the line about trust.

I considered that, but most people who get annoyed about privacy are upset that private companies have any data at all. e.g. location data on Facebook.

It's good opsec to assume all private data at companies may get leaked, including links between your ID and your name.

Consider the scenario where somebody is reviewing books on dangerous subjects (politics, religion, LGBT+, etc...) and is suddenly outed to the whole world due to a data breach.

For some set of people it would be a problem, I think for book reviews that’s a tiny set, not to be dismissive of them, but still.

For other types of applications you’d want to have a better system, like a writing platform.

But still, there’s ways to do it. You can validate the high res copies of whatever you want to validate, then make a hash using a few key numbers, in partial. Stuff like that gets you close to ideal, even the worst case break would expose almost nothing, and you’d prevent duplicate accounts. Only risk is losing the documents during validation before they’re deleted.

I think what actually happens is a new Goodreads minus the technical problems and also minus the reviews and community comes along. Turns out people prefer having actual book reviews to faster page loads and more stable link structures, and everyone stays where they are...

More React will solve everything.

Reply on news.ycombinator.com