Yes so much this.
There are a few issues to consider:
- Any comprehensive archive of your activity is itself going to be a tremendously "interesting" resource for others -- advertisers, law enforcement, business adversaries, and the like. Baking in strong crypto and privacy protections from the start would be exceedingly strongly advised.
- That's also an excellent reason to have this outside the browser, by default, or otherwise sandboxed.
- Back when I was foolish enough to think that making suggestions to Browser Monopoly #1 was remotely useful, I pointed out that the ability to search within the set of pages I currently have open or have visited would be immensely useful. It's (generally) a smaller set than the entire Web, and comprises a set of at least putatively known, familiar, and/or vetted references. I may as well have been writing in Linear A.
- Context of references matters a lot to me. A reason I have a huge number of tabs open, in Firefox, using Tree-Style Tabs, is that the arrangement and relationships between tabs (and windows) is itself significant information. This is of course entirely lost in traditional bookmarks.
- A classification language for categorising documents would be useful. I've been looking at various of these, including the Library of Congress Subject Headings. A way of automatically mapping 1-6 high-probability subjects to a given reference would be good, as well as, of course, tools for mapping between these.
- I've an increasing difference of opinion with the Internet Archive over both the utility and ultimately advisability of saving Web content in precisely the format originally published. Often this is fragile and idiosyncratic. Upconverting to a standardised representation -- say, a strictly semantic, minimal-complexity HTML5, Markdown, or LaTeX, is often superior. Both have their place.
On that last, I've been continuing to play with the suggestion a few days ago for a simplified Washington Post article scrubber, and now have a suite of simple scripts which read both WashPo articles and the homepage, fetching links from the homepage for local viewing. These tend to reduce the total page size to about 3-5% of the original, are easier to read than the source, and are much more robust.
I'm reading HN at the moment from w3m (which means I've got vim as my comment editor, yay!), and have found that passing the source to pandoc and regenerating HTML from that (scrubbing some elements) is actually much preferable, for the homepage. Discussion pages are ... more difficult to process, and the default view in w3m is unpleasant, though vaguely usable.
Upshot: saving a WARC strictly for archival purposes is probably useful, but generating useful formats as noted above would be generally preferable in addition.
With the increasing untenability of mainstream Web design and practices, a Rococco catastrophe of mainstream browsers, the emergence of lightweight and alternative browsers and user-agents (though many based on mainstream rendering engines), the tyranny of the minimum viable user attacking any level of online informational access beyond simple push-stream based consumption, and more, it seems that at the very least there's a strongly favourable environment to rethinking what the Web is and what access methods it should support. Peaks in technological complexity tend to lead to a recapitulation phase in which former, simpler, ideas are resurrected, returned to, and become the basis of further development.