Here is what I would like: I'm typing and I want to cite something. I press a key and I can search by author, title in a database of everything ever published. When I make my choice a cite key is inserted in my format of choice (Pandoc) and the bibliographic information is pulled from that big database to a local file. This file is kept in sync with the big database and my own document, until I decide to 'freeze' it and it becomes static as a permanent record.
Maybe what I want would actually be bad because it assumes that I'm on the internet whenever I write, and doesn't involve a local database of PDFs.
IIRC, ReadCube (https://www.readcube.com/) organizes and extracts metadata of local PDFs and can also search across different search engines to download and extract metadata from online articles.
Not sure how that compares to Zotero, but ReadCube was definitely helpful for me when I was involved in research.
(note: subscription required; offers 30-day trial)
@ipt I remembered that I used the Crossref API to get DOIs for things that were missing them in a script that I haven't thought about for a while.
https://github.com/jdherman/bibby worked for me at some point but now doesn't seem to. It hasn't been updated for four years so maybe it just needs a spring clean.
Anyway, it looks like there are options.
If, also, it interprets copying a section from another paper and pasting it into yours in such a way that it creates a blockquote if the section goes for longer than one line, injects footnotes & citations, and adds the source to the bibliography in the appropriate format, then you've sold me.
I'm glad you're seeing where I'm going for that.
Of course, a full xanadu system is not strictly necessary for that (and probably wouldn't comply with citation standards in its default interface anyhow, being generalizable beyond academia), but it sure would be nice!
All modern forms of hypertext derive from watered-down forms of Xanadu, which is optimized for permanent unbreakable links & quotation. (Even drafts are represented as extensive quotes from previous drafts.)
Unfortunately, it's also famously dogged by management problems. So, while there have been many internal prototypes and some open source releases over the course of the past 60 years, there have been no designated 'full' releases.
An editor like that would be all kinds of amazing! Plus a decentralized database so something screwy wouldn't undermine the whole system.
So, a blockchain-based system? Libraries, universities, etc. could run full nodes, which would also facilitate offline use within those institutions.
A blockchain just means a list where every entry contains a hash of all previous entries. The purpose is to prevent people from forging earlier entries and maintain a consistent order (which makes sense for a financial ledger).
A distributed hash table is what it sounds like -- a hash table spread over multiple machines. If your machine doesn't have what you're looking for, it'll ask peers. A CAN is the same thing but the name is a hash of the content of the document.