I am in the process of writing an image gallery webapp similar to that of Danbooru, and in the process, decided to take a look at danbooru's database schema for an idea about how things are structured. In doing so, I've noticed that the 'uploads' table contains a column named tag_string, which I'm guessing contains comma-separated tags. There's also a tags table which (again, by the looks of it) contains metadata about individual tags.

This got me thinking about the practicality of my initial design whereby I have a M-N lookup table specifically for mapping N individual tags to M different files. My design would use SQL Join to get a result set containing all of the images I'm looking for, while I can forsee danbooru's solution using a search on the text column containing the list of tags associated with that image.

My lookup table could concievably get very large (in terms of # of rows) if every file contains upwards of 20 or more tags, whereas the other solution only extends the size of the files table by one column.

With these differences in mind, what are the pros and cons of the different implementations?

For reference, here is an incomplete (but contains enough information regarding this question) design of my schema.