A trail (IIRC) charts a journey through a set of documents, or perhaps sections of documents.
Think about a case where you are trying to figure out a problem based on a bunch of error messages. You search the web for solution to the first message, then the second message and so on, until you arrive at the root cause and the solution to that. You then save this as a trail (or the memex does this for you automatically, using some fancy ML method).
Now, imagine you encounter a similar or identical issue 6 months down the line. You remember that 6 months ago you searched for a solution, and that you found one. Maybe you forgot to bookmark it. Not a problem, you find the original trail using some simple heuristics (how long ago it was, the keywords from the error message), and then you follow it to the end. Solution found, reasonably quickly.
Yes, and until we have the fancy ML method, one way to do it is to have something like a Google Doc where you document your search by copypasting from the source documents and linking to them.
That was my point with the original article. I'm with you that it would be fantastic to have memex that's described in 'As we may think' (only better), but until we do, I'm doing it the hard/manual way.
I was surprised to recently find out that Google Chrome in fact already has that fancy ML method for chunking history into trails.
Nothing wrong with doing this manually, but I do think that having the Memex do some of the thinking re: how the data is organise would save a lot of time.
