Johnny.Decimal

- It only solves the problem of communicating categories to another person/party with its brevity. 12.03, easy to tell others. But all this at the expense of the following things.

- There is no way to immediately know what 12.03 means. One has to carry a look up table. You can replace it with 2 char code. For example, Finance.Taxes to FN.TX. Why use numbers? Is the ordering important? Why not use char?

- What if categories do overlap? GMail solved this problem in 2005 with labels instead of folders. You can apply multiple tags. Tagging should be enum-like, while tagging it should show what tags already exist so that you dont create two seperate but similar tags, for e.g. Finance and Finances.

- The author dismisses search without giving any reason. Search is amazing for digital documents that is not possible in physical analog documents. The focus should be on tags + keywords or description of the document. Then use something like Algolia search to find stuff - it is extremely forgiving and powerful. Ofcourse, I am just speaking in general about data organization, not specifically about OS-level folders. For that, I think we're stuck with what the OS provides as a search engine. On MacOS, I think Alfred does a better job of searching, haven't used it personally though.

- You can still assign a unique ID to the document, after all thats all the 12.03 scheme does. You can communicate precisely to Jane, "Hey Jane, the document is B75AE2". Jane types that in the search engine and there is no need to weed through folders.

- Additional metadata such as year, author, owner, etc. would help with search.

Instead of making search powerful and contextualizing it, the author expends his arguments on frivolous pursuit of Johnny.Decimal.

Animal asked you to think of an animal, then it tried to identify it by asking you yes/no questions. Let's say it asked you if it has four legs (no), does it have wings (yes), and then it guesses that you are thinking of a hummingbird.

That's wrong, because you were thinking of a penguin. It would then ask you what yes/no question it could have asked to tell a penguin from a hummingbird, and you could tell it to ask "Does it fly?" and the answer is "no" for a penguin.

It would add that question to its decision tree, so if anyone else thought of a penguin it would get it.

For much of my stuff at work, I use an unfiled filing system. I started this one day when I was asked to write some one time report to solve some mystery.

I spent a half hour with the line "mkdir " typed into my terminal window unable to think of a good name for the directory to do this task in. Finally I gave up on naming the thing, and did "mkdir unfiled". Then in unfiled I did "mkdir 1", and "vi directory.txt". In "directory.txt" I added something like this:

"1 Report for Alice and Bob to identify source of unexpected load on main website"

and then went into the "1" directory and wrote the code to gather the data and analyze it and produce that report.

I intended to figure out a real name for the thing later and rename/move it.

I never got around to figuring out a real name, and unfiled/1 is still there. Along with unfiled/2, unfiled/3, ..., unfiled/314.

This has actually worked out well for me. Suppose I run into some kind of issue with UTF-8 handling in MySQL. I remember I've dealt with that before, and grep in directory.txt for terms like mysql and utf. That turns up that the work for those earlier problems is in unfiled/258, and so that's where I go to find the programs I wrote for investigating that kind of issue.

Not everything goes in unfiled, of course. A good rule of thumb is that if it is something my replacement would need to takeover if I got hit by a bus, it should not be in unfiled. It should be in a well named repo on my department's git server.

>Now those same people store all of their files in arbitrarily named folders on their company’s shared drive and wonder why they can’t find anything.

This is the most outrageously false premise on which to base an argument. It's staggering.

When I go through old documents, I don’t even skim through the contents. I just check the date and remove them if necessary.