Amazing stuff, Jan.

From what I could understand, it aims on being a no-brainer reference system, and indeed, my aim is capturing quickly *but* with enough information and structure to be able to find it later. What I find happening to me time after time, is that even though I do have a quick way to capture, I end up not using the information later, because it is hard to find. The reason? I have two reference baskets:
 * The wiki files (under wiki/) -- I press S-w and can type the name of a file and start typing.
 * My GTDReference.org file. C-c C-r r, title and boby, C-c C-c to capture.

(And the GTD Inbox, but that's not reference, only data that I want to process later).

The problem, I think, is that I end up with a bunch of files under wiki/ and a bunch of notes in GTDReference.org. A rgrep search is often enough to find what I want, but it is definetly not as clean, straightfoward and simple as your system's implementation.

In an effort to make the wiki stuff more organic, I have setup howm-mode to work with org, and while it works great to link files, it just added more complexity. I think I will just get rid of it.

Could you share the custom elisp function and the scan shell script?

Also, how does all of this fit on your overall PIM architecture? There's more to it? If so, then I'm ready to get another cup of coffee :)

Thank you very much for sharing all these gems.

Marcelo.


On Fri, Apr 9, 2010 at 6:31 PM, Jan Böcker <jan.boecker@jboecker.de> wrote:
[The following text has gotten quite long. Sit comfortably and get a cup
of your preferred drink if you want to proceed.]

That is an interesting setup you describe there. I had considered
something similar myself, but found it a hassle to come up with a file
name for every new piece of information (although unix does allow
everything except "/" in a file name, I want my file names to be lower
case, short and without spaces where feasible).

To paraphrase what you said, putting things into files just makes you
loose time thinking about file names (which I also consider structure).

In the end, I settled upon one big org file ("reference.org"). Each
piece of reference information is in its own top-level node. When I want
to find something there, I use isearch and/or search for a specific tag
(in virtually every case, a simple isearch for one or two words is
sufficient).

This is way faster than navigating the file system! I also share your
dislike of categories (which a strictly hierarchical file system would
force me to use).

I have two remember templates to add an entry to reference.org.

The first template asks me for tags ("%^g") and a title for the
headline. After filing it (at the top of reference.org) with C-c C-c,
Org jumps to the location it was just filed ("%&" in the template), in
case I want to use C-c C-c again to readjust the tags.

I use this first template to keep data that can be expressed in plain
text (including all the powerful tools Org gives me to work with plain
text, such as outlines, links and tables).

The second template is a little more complex; it calls a custom elisp
function to do all the work.

When I call this second template, I am asked for the following:
- a title for the headline
- a date (I mostly use this template to file scanned letters, invoices
and the like, so it helps to be able to change the date from the default
of today.)
- a folder name (defaulting to YYYY-MM-DD.S, i.e. the previously
specified date followed by a sequence number to make the folder name
unique). Normally, I do not customize the folder name, because I only
need to find the reference data via Org and do not need to navigate to
it using e.g. the "open" dialog of any other program (and I do not want
to change it in the future, which might make a folder name containing a
date obsolete).
- where the original went (defaults to "Trash"). This is stored as an
attribute in the outline node.

A new subfolder with the specified folder name is created in ~/org/data/
and set up as this node's attachment directory. The ID of the node is
set to data-<folder name>, so I can link conveniently to this entry from
project notes.

The custom elisp function also installs a hook that automatically calls
org-attach-attach-mv if I try to file the template without having added
any attachment, so I do not forget this.

I use this second template when I have to attach a file to the entry,
because it cannot be represented in Org. This mostly applies to scanned
paper of any sort (letters, invoices, etc).

I have a shell script which I use to scan directly to PDF files (I do
not use OCR, the PDF just serves as a container for possibly multiple
scanned pages, so that browsing and printing the whole document is
convenient).

If it was an important document where I might need the original in the
future, I specify "Filed" when asked where the original went, write the
folder name/ID number in the top right corner of the document with a
pencil (in case of very important documents or certificates I use a
post-it note), and file it on top of a normal paper file folder.
When keeping the original, I do not change the folder name from its
default. Should I have to dig out the original for any reason, I can
manually execute a binary search on my  chronologically sorted file
folder(s).

I have been using this system for a few weeks now, and it has worked
great so far. Its main design goal was not simplicity of implementation,
but simplicity of use: it has to be so simple (and require as few
decisions as possible) to file something that I actually do it instead
of postponing it.

The system actually evolved along with the aforementioned shell script
for scanning while I was scanning and filing about 20 exercise sheets
(about four to twelve pages each) from the last semester to access them
conveniently when preparing for the exams.

I also noticed a while ago that very long org files become less
intimidating once you learn to love C-x n s (org-narrow-to-subtree),
which helped with my decision for one big file over many small ones. One
big file also avoids cluttering the buffer list.

- Jan,
 who really should start a blog to do more detailed write-ups of this
and similar things, because they are so much fun to write.