Comment by mikewarot | Hacker News Reader

Comment by mikewarot | Hacker News Reader

Parent Full thread

There were 3 main features of the Memex that we just don't have.

Annotation - The ability to mark up hypertext... yeah... HTML claims to do it, but you can't change a document once it's read only. You should be able to layer annotations on top of a document. HTML simply is not a Language for the Markup of Hypertext Documents.

[Edit/Extend] - Annotation is anything you could do with a printout, at minimum. Circle something, highlight a section, black something out, attach links, etc. It's not just tagging a file as a black box.

Trails - the ability to store a trail of annotations, linking up parts of two documents together, or more. Like a stack of bookmarks, except bookmarks can't anchor to part of a document.

Publication - One of the crucial features of the Memex was the ability to output a collection to allow sharing with others. It could spit out a copy of everything linked to a trail, all the files, all in one dump. Copyright holders would have a huge problem with Memex if it were implemented as originally shown.

I think it should be possible to build a proxy that records ALL local http(s) traffic, and thus record all the sources to documents you've seen... and then only keep those that you've tagged or linked to after a few months, purging the rest. A private Memex, but it couldn't even be an open source project, as copyright holders would work to squash it.

Oh, we did all that. I even wrote my PhD dissertation on it (2001). Back then, there were quite a few research projects working on what I then termed ‘Web Augmentation’. You could do it serverside, in special proxies, or in the browser through extensions or scripting. The earliest browser based system I know of that allowed users to insert links and annotations into existing Web pages was DHM/WWW, developed by yours truly in 1997 – we even won best paper for it at Hypertext 1997. This was a specialised case of what was then known as Open Hypermedia, i.e., a hypermedia system that could integrate (more or less) arbitrary documents and systems through links and other structures.

Using a more general architecture, it was possible to add many kinds of structures on top of existing Web pages collaboratively. Navigational or spatial hypermedia, guided tours, etc. Link bases were stored outside of the linked resources, so that you could have different sets of structures applied to a corpus of documents.

There were also quite a few commercial attempts in this space. However, all, commercial and research alike, ultimately failed, and I believe that there were several reasons behind this: Web site owners disliked that others could annotate their pages; shared link bases often became polluted; bookmarks were ‘good enough’ for most users; maintaining link anchor consistency in (increasingly dynamic) Web pages became more and more difficult (despite our best efforts); and while I remain enthused about the notion of the MEMEX scholar carefully annotating their corpus of literature, I don’t think the appeal is all that widespread.

Still, there are remnants in some systems – e.g., I enjoy to see that others have highlighted the same passage as I in a Kindle book.

> Annotation / Edit / Extend

It's weird — we, the builders of HTTP-based APIs, very well understand the concept of "REpresentational State Transfer" ...but I've never seen a web-app server that actually accepts PUT or PATCH or POST of HTML Content-Type against the baked-out server-rendered HTML views, such that if you were to push a modified version of said representational state (the post-edit HTML), it would get "decompiled" on the backend into a delta of the internal state (e.g. Markdown in a CMS), and that change to the internal state would get persisted, reflecting back as a change in the representational state when the page is fetched.

If anyone would just actually implement that on a web-app backend, then you'd be able to "annotate, edit, and extend" HTML-presented content using freakin' NSCA Mosaic from 1993. Or the "edit in Frontpage" feature of IE4.

(No modern equivalents, though — the features that would generate a "native" browser PUT or PATCH against an HTML page have since shrivelled up and died, after it became clear that no server would ever bother to allow them.)

To dodge the chicken-and-egg problem of browser support, you could do this today, with a Javscript SPA serving the browser role: the frontend app would ensure that it's reflecting all internal state changes in its in-memory client-side internal state into changes to the HTML (even if those changes don't cause any presentation changes on the client side); and then, where a regular SPA would push those changes by calling an API endpoint that accepts a version of its in-memory client-side internal state, this SPA would just push the updated HTML, as a PUT or PATCH to the very same URL that was loaded to deliver the SPA.

I'm not sure about others but I'd prefer my extended brain not live on other people's servers. The Memex always seemed to me to be a personal archive and notebook. You would pull in material from outside and store it locally along with your links and annotations and then only publish the work that you wanted to make public.

This might have been a “Missed Connection” for micro formats. We could/should have been able to PATCH those parts.

Someone at NSCA implemented annotations in Mosaic, but the bus number dropped to zero when they left and the available evidence seems to be that none of those of us after the exodus were familiar with Memex. I think a coworker recommended Vannevar Bush to me around 2006. Boy was that a Tyler Darden moment.

I think what the world has needed basically since then is browser integration for something like annotations, so that web sites and commentary about them can live together. And with less of a direct line to advertising dollars. Perhaps something that could live in the Fediverse. Sometimes I want HN’s opinion and sometimes I want someone else’s.

> I think it should be possible to build a proxy that records ALL local http(s) traffic

There was a Microsoft Research project in the early 00’s called “Stuff I’ve Seen” that did basically this. I don’t know why it never really went anywhere - I’ve wanted one ever since I saw the talk.

Edit: Link - https://www.microsoft.com/en-us/research/publication/stuff-i...

I think, the Permaweb comes pretty close.

You can annotate any upload (arbitrary data, files, or collections of files) with tags and since the data is permanent, your links don't break.

Publication is possible too, and with the Universal Data License, copyright holders can define the terms for their part of the publication right at upload time.

I would think that Apple has the level of integration that would make this possible, but then... where's the potential profit that would make it worth their while ?

If users found value and bought Apple devices because of it?

Apple has lots of software that only exists to sell hardware. This seems compatible with that business model to me.

Point taken. But things sold by Apple must have an immediate, easy-to-digest appeal. Is it possible here ?

Reply on news.ycombinator.com