[Thomas Ryan]

Originally Posted by Krystian Galaj

With folder named A, containing subfolder named B which contains subfolder named C, and subfolder named C which contains subfolder named B:

A -> B -> C
   -> C -> B

how do you distinguish files in B->C subfolders from files in C->B subfolders?

I think you might be thinking "navigation" vs "underlying data representation"

Consider -

A - my books
B - fiction/non-fiction
C- Author X, Y or Z

folder content:

mybooks -> fiction -> X-author -> book_id_1
mybooks -> fiction -> Y-author -> book_id_2
mybooks -> fiction -> Z-author -> book_id_3

mybooks -> non-fiction -> X-author -> book_id_4
mybooks -> non-fiction -> Y-author -> book_id_5
mybooks -> non-fiction -> Z-author -> book_id_6

(where book_id_x, can be a single book, or a list of books)

tags:

book_id_1 = {mybooks, fiction, X-author}
book_id_2 = {mybooks, fiction, Y-author}
book_id_3 = {mybooks, fiction, Z-author}
book_id_4 = {mybooks, non-fiction, X-author}
book_id_5 = {mybooks, non-fiction, Y-author}
book_id_6 = {mybooks, non-fiction, Z-author}

inverted:

mybooks = {book_id_1, book_id_2, book_id_3, book_id_4, book_id_5, book_id_6}
fiction = {book_id_1, book_id_2, book_id_3}
non-fiction = {book_id_4, book_id_5, book_id_6}
X-author = {book_id_1, book_id_4}
Y-author = {book_id_2, book_id_5}
Z-author = {book_id_3, book_id_6}

compare:

A->B->C with A->C->B

for some example A= "my books", B = "fiction", C="Y_author"

A->B->C = {book_id_1, book_id_2, book_id_3, book_id_4, book_id_5, book_id_6} -> {book_id_1, book_id_2, book_id_3} -> {book_id_2}


A->C->B = {book_id_1, book_id_2, book_id_3, book_id_4, book_id_5, book_id_6} -> {book_id_2, book_id_5} -> {book_id_2}

The "And" operator is commutative, so the end result is the same. I agree there are two different paths there.

So in a folder dominated UI, you might have a different navigation experience to find your target than the more general approach. That seems irrelevant given the next observation.

e.g. in this case - given no priorites/levels in the tags - I choose one tag, let's say "mybooks", I am now presented with a second selection choice from {fiction, non-fiction, Author X, Author_Y , Author_Z}. I pick one, and the choice refines.
I am not forced to navigate according to a single folder scheme. (your mental model vs. mine)

Note the power of how this generalizes - say "A" extends to be {"my books", "your books"} - I can find a particular book_id starting with my books vs. yours, or I can start with author, refine by fiction-type, and then add the ownership criteria at the very end so I know which owner (library?) to go to.

Someone pointed out that the flattened list (second selection above) may be too big, but lists are too big all the time, and that is an entirely separate issue. On a huge screen less a problem than on a cell phone regardless of approach. I would like to separate presentation from the underlying representation. I am happy to understand what I am missing on this topic.

Last - someone mentioned "ontology". I am not sure what that means in this context, but I will agree that with a pure "tag" approach what is lost is the "category" name for a selected subset group of tags. In the example above there is no wording for "fiction value". So, unless this data is stored elsewhere there is no way to label a certain groups of tags. But who cares? (rhetorical) Make a more flexible User Interaction without that constraint. Live free from folders! (Or, if you insist, have tags for tags; ha! think that one through)