Re: One vs many directories

[Jean Louis <bugs@gnu.support>]

* Dr. Arne Babenhauserheide <arne_bab@web.de> [2020-11-22 01:48]:
> > So in general I never need to use some general search through Org
> > files or any other files as my way of thinking begins with People or
> > Groups and that narrows what has to be searched.
> 
> How do you deal with stuff that applies to several people?

From database viewpoint there are

- accounts (which means companies, groups, entities, like "People who
  wish to get employed in Europe")

- there are contacts, that may belong to account, additionally belong
  to company (also account), additionally be member of account, so
  there are 3 groupings for each contact how that contact may be
  related to account. If it is main account such as "Welders" or if
  maybe under "Company" is written "welders" (not quite correct) in
  reality it does not matter.

- then there are lists to which other lists belong. Account A and
  account B, C, D can belong to list 01. Various accounts can be put
  together in uniting lists. Those lists are encompassing other lists,
  not individual people but people in the list (account) usually
  unless there is only one in the account. Those lists I am using for
  mailing them or informing them by letter, SMS, etc. Geologists and
  mining engineers and metallurgists are 3 different accounts but if
  all of them speak Swahili both in Kenya and Tanzania and are in the
  related branch of economy so they can be sent same type of
  information.

Then there are groups, which is just another name for a new list. Then
there are tags. I can freely tag account, contact or anything else. By
tags I can finely select specific people belonging to specific group.

There are account types and group types.

Tags by itself have its own description or purpose to name it type.

Some people introduce other people, few of them introduced
thousands. So contacts have a column "introduced by". That becomes
very handy when talking to somebody and it also helps in awarding
introduces. It helps when people place their hyperlinks and become
automated introducers (lead generation).

When I know that person belongs to some group of people and I have to
write email and I know it is better to inform everybody, then there is
action to assign identity from which I am sending email. It could be
public or internal identity with different email address. Emails to
that person would always go from designated email address without me
thinking about it. Then there are Cc and Bcc fields and in those
fields I could designate: to inform same contact to each of email
addresses with same message, and to inform group of people each
time. Thus if writing to one contact all others get informed through
same message. But I am not thinking about who belongs to the group,
and what are their email addresses, that gets inserted
automatically. Email composition is usually inside of Emacs.

Sending files? If contact is in the group and others need to see the
file, then Cc and Bcc fields work in the same way, file sent to one
person is sent to other members of the group.

Sometimes contact tells me to please exclude some people from Cc, so I
just go into definition of identities for contact and exclude such.

SMS I am sending by using kdeconnect.el package so any SMS I send this
way it is getting recorded to the contact. If I need to send SMS to
the group, then same SMS could be sent to whole group.

If the note on one contact is related to other people then is trivial
to automate to inform other people on that particular note.

If any group of people is there, filing files under group does not
make much sense to me. I would not file it as Group/ABC/file-01.pdf

I would file it under By-ID/Contact-ID-123/Group/file-01.pdf as files
are normally sent by one email address of one person or under one
name. I deal with people not with empty groups that do not
communicate. But group is important, so there can be /Group/ directory
on the file system where all contacts of one group can be symlinked
automatically if it happens that I need to search by Group on file
system, but I don't. I search for groups in the database and see list
of contacts there and then jump to contact.

Same thing with invoices, they are written either to company or to
individual or some individual in some company. I will not file or act
based on invoice. I have to act based on authorirative or maybe
hierarchically higher object first.

In general is always good to make a list of things and then lists are
foundation for dealing easier with whatever groups of anything.

> > it comfortable. My way of thinking is always People or Groups, and
> > from there various searches are performed and that narrows drastically
> > the subject that has to be searched.
> 
> That does sound like it should speed up searching by directory.

You remember that I do not search by directory.

Computer stuff I search by directory, like
Work/Computer/Devices/Dictaphone or Work/Computer/Graphics/shadow

Stuff related to any entity, organization, group of people, mailing
list or list that encompasses other lists, I am locating in the
database but the file system location is automatically expanded from
the unique IDs. So I need not know where is exactly directory for the
group ABC, but I know if I press F8 that I am right there. Then files
are sorted for the group or account normally by date, but could be by
License number or something else. It is mixture of database relational
jump to base directory for specific account/company/group and
sub-directories. Base directory I do not need to know, those
sub-directories are few and search becomes very fast. No special
database is needed for this as one can keep hundreds of groups and
people in simple Lisp structures saved in files.

> My mind works differently here: I remember some concept and need to
> find stuff connected to that, including people, but also additional
> ideas, instructions, and just bullet points with info I might need
> again later (which multiple times saved many hours of searching
> already).

What I learn is that concept of management of information in computer
better works by the concept of associations or relations as that is
similar how mind works. I do not see much difference here in thinking
how things information should be stored.

I have some hobby and there are paper books, images, videos, DVDs,
digital books related to that hobby. All that is sorted by Authors as
that is one attribute that those things have. Additionally there are
subjects. Authors's individuals works are symlinked to various
multiple subjects as multiple subjects (tags) related to same Author's
work. If author produces only for one subject then his name is
symlinked to that subject. Sometimes there are 2 or more authors
working together, so their directory is also symlinked to individual
authors.

Similar concept applies to music or videos, it is sorted by authors
and bands, and then individual works are sorted by its type of music
or genres of videos.

Images are sorted by people who are on the image. But I do not think
of that, I just say "Adrian" and image is sorted there in appropriate
date folder for that person. Computer thinks WHERE is the location,
not me. All other files are also being sorted like that and it is work
in progress.

> The one thing that keeps me from that is that I often file under
> specific projects, and the active projects are shifting
> constantly. For that I enjoy it a lot that I only need to customize
> the capture templates to add a project.

In the dynamical knowledge repository backed by PostgreSQL I have
subtrees and each subtree has its name. I am more familiar with
structured data in the database. So if I wish to have equivalent of
capture it becomes trivial, press key, choose set and insert whatever
note or task or article, anything. It just has little different
structures.

I do not need to setup extra capture configuration as there are
already sets or subtree names. I am quickly locating appropriate
substree name and filing under.

Something I wish to file need not be Org mode, but it can be. A note
can be written in any text mode and filed in the set or
subtree. Markdown, Koutliner, it can be Perl snippet, one liner,
excerpt from email or message mode note, why not. SQL snippet, Emacs
Lisp or Scheme. Normal text. I feel more free this way. 

Hyperdocuments have its types like WWW, Gopher, Gemini, FTP, SFTP,
Note, Task, Video to be played at exact time, local file, YouTube
video, Dired directory, Program Launch, PDF, PDF by query, PDF by page
number, Emacs Lisp, Org Heading, Org file, Message ID, Set, and so
on.

These days I am thinking to make self-defining types and I mean to
place the code inside of the hyperdocument type entry that defines how
that entry should be handled. User accessing database at University
Makerere could access hyperdocument that designates that not only
video file has to be opened in the browser, but that window has to be
split so that annotations related to video file has to be shown in
same time. But user accessing such hyperdocument need not necessarily
know that its viewing or launching is customized by the type of the
hyperdocument defined somewhere else, not in software directly. It
sounds similar to Javascript launched in web browsers with full
liberty in computing. Emacs Lisp snippets would be pulled from
database directly and executed on users' computer. File transfers are
non existent and file system is not there. Nothing gets really
downloaded or saved, just loaded into memory and executed.

Just thinking of a hyperdocument type or object type "comment" or
"feedback". When activating such user loads few functions, write the
comment, functions can also say what will be done with the
comment. School teacher or head teacher acts upon it or comments are
simply published.

Sets for now do not have types, but why not. If set has type of topics
and topics allow comments there is automatic forum and discussion.

> I for one am still working on my workflow, and I guess that 10 years
> from now it won’t be the same as today. I hope that learning about other
> ways to work with org will help me a lot in future.

Same on my side.

Summary for me:

- navigation by hierarchy is one way, we all use it to search and file
  documents
  
- searching through indexed database is different way, it is not good
  for filing anything

- direct relational access where computer locates the file is third
  way for knowledge management

- it is better designing filing, sorting and retrievals by concepts of
  mind or how mind works

- any objects or pieces of information shall have its actions that act
  upon it and that can hyperpush the user to other pieces of
  information

- general increase of hyperlinking by relations helps in creation of
  better systems.

Here is how I have implemented simple versioning system that is
decided by program and not the database. It was just before hour.

(defun hyperscope-vc (table column id &optional description)
  "Simple version system."
  (let* ((value (rcd-db-get-entry table column id *hs*)) ;; takes DB value
	 (value (sql-escape-string value)) ;; prepares value for SQL
	 (description (if description (sql-escape-string description) "NULL"))
	 (sql (format "INSERT INTO vc (vc_table, 
                                      vc_column, 
                                      vc_tableid, 
                                      vc_value, 
                                      vc_description) VALUES ('%s', '%s', %s, %s, %s)
                         RETURNING vc_id"
		      table column id value description))
	 (id (rcd-sql sql *hs*)))
    (if id id nil)))

Above function takes the value from any table, any column, by ID and
inserts the value into the `vc` table. 

Then comes this function below that edits hyperdocument as Org file
with Org mode. Before updating it is using `hyperscope-vc` version
function to basically save the old value in the separate table. If I
want to see diffs I think it is also trivial to do.

(defun hlink-edit-org (id)
  (let* ((blob (hlink-description-value id))
	 (blob (if blob blob  ""))
	 (buffer-name (format "HyperScope Editing ID: %d" id))
	 (new-blob (read-from-buffer blob buffer-name 'org-mode)))
    (hyperscope-vc "hlinks" "hlinks_description" id)
    (rcd-db-update-entry "hlinks" "hlinks_description" "text" id new-blob *hs*)))