I've just started speccing out some thoughts on how document references (files, URLs, URIs, ...?) might be improved. A sort of api / multiple-attribute interface strikes me as useful:
1. Meaningful titles / filenames.
2. A content hash. Not exactly human-friendly... I'd prefer something shorter to longer, though there's the collisions problem. Still thinking on this. Could tie into VCS (git, Hg, etc.)
3. Various standard metadata: author(s), editor(s), publisher(s), and their relation(s) to the doc. See various bibliographic or MARC 21 formats.
4. Dates: Initiated, published, modified, accessed, read date(s).
5. Classification schema. Here with time I find that working with extant rather than de novo or ad hoc schema is most likely preferable. Dewey Decimal is apparently more logical than LoCCS, but the Library of Congress classification is nonproprietary. That may well be what makes it win out.
6. Workflow indicators. Unread, read, deep-archive, to-process, etc.
7. Reputation and ratings. Associated with works, authors, other contributors (editors, fact-checkers, etc.), publishers. Some form of distributed assessment here would be useful.
The question is how to make these visible. My thought is to treat the parameters as, essentially, search. It doesn't matter what name you use, so long as the set characteristics is unique or determinable. E.g., returning 3 items would allow me to manually determine which I'm interested in, returning 300 would make that difficult, returning 3,000 would probably require some programmatic determination (or further specification).
A URL scheme, say, URI:/au=clarke&dat=lt:1980&ti:=imperial&ti=earth ... could, say, return Arthur C. Clarke's Imperial Earth.
A filesystem / virtual filesystem approach might work somewhat similarly locally, either from a CLI or graphical approach. I'm thinking of how that might work....
Very early stages.
My view is pretty much as yours: POCKET IS MY VETTED WEB. If I've saved stuff there, it's because that is shit I've already determined is worth a second look, and I want to be able to find it again.
Increasingly, when I hit my Pocket archive (3k-4k+ items, top 1% user), I find that I cannot find what I'm looking for. This negates approximately 80% of the purpose of having Pocket in the first place, and I've been giving strong thoughts to simply using it as a link dump until I can figure out how to:
1. Download the content myself.
2. Download the tags, which are the most important element.
3. Get the simplified-Web view. Web design isn't the solution, Web design is the problem.
4. Be able to search and filter specifically what I've already fetched. And filter by more options: date, website, author, tags (positive and negative filters), etc.
None of that currently exists. And it is increasingly frustrating.