The same people who killed folders "because tags". Don't get me wrong, I love tags, but why on earth can't we have BOTH???? I don't always remember what I tagged something with, and searching through 8 years of work to find something is useless.
Tags would be okay as a folders replacement if they, you know, actually replaced folders. That is: if a tag can in turn have tags, and those tags automatically apply to the tagged files, then it should be trivial to not only emulate a folder hierarchy, but even improve upon it by offering more flexibility.
It'd also help substantially if more tag-centric applications were up-front about which tags are already defined.
Very few classification schemes are unambiguously DAGs.
We're used to them for use with physical objects, which can only occupy a single space. When the search interface is the representation (as with files or data stores), cross-referencing across categories is readily facilitated, and (in the extremely rare instance of well-designed ontologies), useful.
I've created several collections using extensive tagging schemas, and not infrequently run out of tags. I've also been looking at various ontologies suggested over the years, dating to Aristotle, Francis Bacon, Diderot, and more recent (often library or encyclopedia) schemas. Some are useful, none are perfect, all incorporate biases and anacronisms.
Another feature of folders (directories), or physical storage locations (for physical items), is that the classification / relation is created automatically by virtue of where an item or file is located. For tagging systems, you've got the additional overhead of specifying (or creating a rules-based tag-generating system) to apply tags to content. Otherwise you end up with a morass of "My Document0012811.docx", "untitled", "unclassified", "misc", etc., content. Cataloguing itself is a role, process, and step, and requires maintenance. Even within library science, the tendency has been to move toward self-describing works, full-text search, and imputed relationships.
Sure, but that doesn’t explain how you magically remember where you stored the file. I suppose it does help with making you unable to store files in the root directory and forcing you to store it somewhere with some meaning. This seems like it would just resolve into a documents directory and you’re back where you started with a flat bag of files that you need to organize arbitrarily.
Also, you can use hierarchies with tags too, it’s just a matter of interface. Example: click “photos”, and see all the other tags that are associated with photos, and in a separate view all the files with all the previously clicked tags. You can drill down this way. Order top level tags by number of files using it. I have never seen this implemented but it seems like it would be fun. I also imagine “photos” itself would be a special file type metadata tag.
EDIT: posted before fleshing out idea, it is fleshed out now.
Great argument for tags! Sure, it's never been done, but it's possible that someone will re-implement the inherent properties of filesystems with tags in the future, so let's take your filesystem away now!
Hey I am just describing interfaces, I don’t care about how it hits the disk. Filesystems are certainly nothing to get excited about; links are a dirty hack to get around a forced tree anyway.
