iAnnotate – Whatever happened to the web as an annotation system? (2013)
scholarlykitchen.sspnet.org
scholarlykitchen.sspnet.org
In 1995 Dan Huttenlocher and I created CoNote, which supported annotation on web documents. See Computer Support for Cooperative Learning, 1995 conference, http://jamesrdavis.org/Papers/CSCL95/cscl95.rtf
Sadly, it's safe to say it never quite took off, and the only collaboration effort that has seen some use in quantities remains the much more modest pingback/linkback protocol. It's not clear to me, though, whether pingback is/was only popular because it could be exploited for DDoS attacks.
* Ease of use: Getting someone to install an extension is really hard (and doesn't even work on mobile for most users), and getting them to use a specialized browser app is even harder. Firefox would be in a good position to do something on this front. Conceptually, it could even fit in with Pocket. (e.g. "Pocket saves content from everywhere, and allows you to make comments on anything.")
* Canonical addresses: A page/topic/article story can exist at many slightly different urls. How do you properly/automatically consolidate comments? The problem becomes adversarial as soon as it gains any bit of success. No company wants competitors to have the ability to communicate with their site visitors, and will do everything they can to keep comments from being discoverable.
* Spam/moderation: Lots of options here, but none I know of that don't involve big tradeoffs between cost and usefulness.
The good news is that there's probably some opportunity for an "AI" assistant to serve much of the same function as a web-wide commenting/annotations system. The bad news is that if it comes from any of the major players, it might be too sanitized/milquetoast to be useful.
In my experience these browser plugins have come and gone likely due to being more an exploratory coding project rather than a multi dev community supported endeavour.
No idea if the real demand is there tbh.
Another thought is that such a comments platform already exists: bitcoin. Anyone can attach a message to a hash associated with a url (for a fee). It's gonna be awfully slow, though. And since it's global, any popular url will be buried under mountains of trash, including stuff that will make govs shut down bitcoin completely.
An AI system that tries to summarize the annotations to tame the chaos might be interesting, though whether people are willing to "engage" with a system that takes their input and merely feeds it into an averaging model such that nobody will ever see their actual comment might be a tough sell.
I haven't seen or heard much about the development or adoption of their web annotation concept since that announcement...
I'm not exactly sure why the annotation idea is bad, but it has been tried many times now (yours truly included) and has failed over and over again. It probably has something to do with our faulty intuition. It makes sense that it should work (the internet is virtually the perfect medium), but in reality, it stubbornly doesn't want to.
[1] https://variety.com/2021/digital/news/genius-sold-medialab-w...
Making that opinion into something that others would read, and not only for lulz, is a different, much harder trick.
And, of course, commercial.spam and non-profit vandalism are problems every commenting system must address.
It's definitely proved to be a bad business multiple times over, but I foolishly hold on to hope that it's a good a good idea. (Wikipedia might have some parallels.) Without a really low friction way to discover and contribute comments, it's definitely not going anywhere.
Nobody cares about annotating everything (except geeks like me), but people working within specific documentation domains require annotation as part of their workflow.
This meant that using that browser passed through squid and all the web traffic was cataloged. The browsing could then be annotated and cataloged for the office.
It's the kind of thing that takes a while to get configured but after that it is easy to replicate as part of the standard deployment.
Anybody could use any browser to do what they want without passing through the proxy server. One browser was configured to use it.
(How big is the demand to read annotations? What sort of annotations?)
He was trying to get factual feedback on climate change news, a bit like Twitter community notes now.
Built a web annotation system to enable that.
But, instead of commenting on a URL, what if you could comment on a snapshot in the Wayback Machine on The Internet Archive by using that link instead. And then make a hash of that link and post your comments (and search for others' comments) on an open platform like NOSTR?
The building blocks already exist today.
Here is an example: https://snort.social/e/nevent1qqsytrx2ya4854mvhz5tvpwzutvdlf...
Just prepend smort.io/ before any URL to read.
It also works on ArXiv papers published till May 2023.
[1] https://smort.io
Would there be a benefit in social network features? E.g. I would love to see the currently most shared and annotated articles for any topic. It would also be interesting to see the Tweets and posts that reference an article.
The commercial web shifted the development from people self-publishing to ads that needed to be sold.
This has had a tremendous effect on everything.
The best example I have is the massive infrastructure many people insist is required for a website.
This is my biggest pet peeve, and I think a lot of it (among supposedly tech-savvy people at least, less technical people are a different story) is caused by people looking at the cost of a random selection of AWS products, often quoting on-demand prices rather than the 40% discount you can get by buying a year of reserved capacity at once, multiplying by 12, and then freaking out.
Many cloud products are not good deals, and almost seem designed to make people think running a web service is inaccessible to them. Those products are usually given a healthy markup because you’re paying to avoid certain setup steps or for the ability to scale infinitely large in two clicks.
You can still just rent a few cheap servers (or even just one) and, if you set them up properly, you can run a decently sized website off of them no problem.
It's something that's on my mind when I think about launching a site that's intended to draw a significant userbase. Back in the day I'd set up VPS instances with nginx+unicorn+rails and it was relatively smooth, but security has seemingly become so much more critical that I don't know I'd trust myself to get all the biggest holes patched up and more importantly, keep them patched.
Also: people use big services for discovery. If you write a blog, nobody's going to read it unless you get out there on the social media and promote it.
How important is it to discover a blog post no one is talking about?
A writer needing an audience is nothing new. I think it is just as valid to create no matter what comes.
A personal blog is a website.
A forum is a website.
YouTube is a website.
Amazon.com is a website.
Twitter is a website.
They all require very different technology (especially backed) very different technical resources, vastly different amount of labor.
Also, on a lot of smaller websites, crawling bots produce the majority of visits. With www being so big, and its common attractors so strong, it's really hard to have an audience.
Since code is also just an asset on the Permaweb, you can even define renderers, so every file comes with a web app that can open it.