ID	33e2b9f5-3a7e-478b-a270-8a2a2e9e3a75
DeertopiaVisibility	public

Tag-oriented organisation

META: This file is used to discuss tags in the abstract, without thinking about the potential connection to a file system.

Computed tags

ID	ac7e98a9-110f-43b5-b82c-a000d86b49ea
DeertopiaVisibility	public
ROAM_EXCLUDE	t

Tags that can be computed from other factors might be wasteful to store on-disk. Imagine a tag mp4 that is always present if the file has the extension .mp4.

Could rules like these could be user-programmable via a Scheme/Lua DSL? If so, how do we keep them fast?

Some computed tags

For source repos, Github-style language breakdowns.

Complex indirect tags

Nayuki's article proposes tag cores to address the points where "simple tags" fail. To summarise the issues described in the relevent section:

Renaming a tag can has destructive potential.
Tags are identified by a single string, making them very sensitive to language; language as in English, or Chinese, but also as in synonyms and alternative spellings, such as file system, filesystem, filesystems...
Ambiguity: Does the tag apple refer to the company, or the fruit?

Most interestingly, these are problems that simple tags were intended to solve! While we have solved many of these problems on the file-level, we now have to deal with them on the tag-level! I can't help but draw analogue to type theory: Many issues solved by types apply to the types themselves, thus leading to dependent type theory. Let's follow that path, and get meta.

Let a tag be a pair (name : Text, target : FileID)

Automatic tagging

Related to Computed tags. New files should be automatically tagged in a variety of ways.

Some automatically-applied tags

created by: APPLICATION-NAME — By which application was this file created? Firefox? OBS? bash?
owner: USER — To which user does this file belong?
date of creation: TIMESTAMP — When was this file created?

Application-specific cases

Tag-aware applications can greatly improve the utility of automatically-applied tags.

Firefox

online url: URL — Where on the internet does this file orginate?

OBS

duration: DURATION — How long is the recording?
recording: {Team Fortress 2, Desktop, ...} — What was being recorded?

Tag implications

While tag implications allow emulation of directories, they're actually more expressive! Like directories, you can draw trees:

digraph tree {
  bgcolor="transparent"
   node [
    style=filled
    fillcolor=gray95
  ]
  Documents [ shape="record" ]
  Books [ shape="record" ]
  Papers [ shape="record" ]
  Documents -> { "Scheme R6RS spec.pdf", "Emacs reference card.pdf", Books, Papers }
  Books -> { "Dummit and Foote.pdf", "Compiling with continuations.pdf" }
  Papers -> { "Why Functional Programming Matters.pdf", "Dependent Optics.pdf" }
}

file:tag-implication-tree.png

Unlike directories (without {hard,sym}links), you can draw diamonds:

digraph tree {
  bgcolor="transparent"
   node [
    style=filled
    fillcolor=gray95
  ]
  "TODO: example lmfao. i'm going back to playing factorio"
}

file:tag-implication-diamond.png

References

[cite:@nayuki2017designing]