Art of Agreement
Back in 1960ies, Andrzej Trybulec told me that the immediate memory of an ordinary person can grasp of hand not more than seven different items. That's why, as a rule, I have no more than 7 children under any graph node, when the graph is meant to be remembered and understood by a human (as oppossed to a computer). Of course a node may have a homogenous array of children. For instance, a parent node P may include children
a000 a001 ... a999
But if node P has also other children beside say a000...a999 then the tree should reorganized. Say that the children are
a000 a001 ... a999 B C D
Then one should create a node A which has a000...a999 for its children, and node P should have
A B C D
as its children; i.e. now a000...a999 become grandchildren of P.
My example below will serve as a methaphor for much general situations. That's how ancient Chinese would teach. They provided generic examples.
I also use the notion of hard drive (or hard disk) symbolically (not literally). In particular there is an intuitive distinction between so called hardware and software constructions (we talk about these two notions even withinn software; some software constructions are hard relatively to other software constructions).
So, here is my example. I want the system of my files, say for Everything 2014++, to be perfect. On the other hand the system evolves. It's not realistic to count on predicting everything in advance. And mistakes and corrections are likely to happen too. It follows that the hard system on a hard disk cannot be perfect hence such a system should not attempt to be perfect. Let such a system grow evolutionary in the hard disk just like so many things in the nature, including natural languages, etc. Thus, for all the practical reasons, the ideal that related files should be in the same directory cannot happen.
Actually, there is a fundamental reason for an impossibility of a perfect organization--in the complex situations a perfect organization simply does not exist, period. For instance, mathematicians attempt and often succeed in writting very logical mathematical textbooks. Beautiful. But such a perfect or nearly perfect organization is impossible for the whole mathematics. Such perfectness is a nonsense. The same goes for human society. The best the humans can do is to obey the non-imposition command.
Now that we are uneasy about a perfect complex system of files on our hard drive, we know the solution: we can introduce an organization away from hardware, e.g. by introducing the subject index files. In terms of computer science we can talk about indirectness or about indirect addressing. This way there is no need for ordinary users to know the hardware file system. All that ordinary users have to know is the subject index files.
The topic of the actual hard organization is still important but it's a different theme.
Once again I will present a concrete example (ok, pseudo-example). The names of the nodes in real life can be quite arbitrary. It's to the advantage of a user to be presented preferably with full names rather than abbreviations. Thus a subject index can be difficult to view as a single file. It seems practical to present segments only, one at the tiime. Let say that T is the top node, A B C D are children of T, then Aa Ab Ac are the children of A, and the rest od the system looks similar. Then the subject index may consist of several files where the top one looks like this:
I will consider representing a complex system. For the sake of simplicity I will still talk about the system--I don't want to use the cumbersome term representation of a system.
It is common to partition a complex structure into disjoint components. Here however I will provide an example of a cover rather than of a partition, i.e. the components will overlap. The overlap will be relatively small hence the cost of introducing a cover relatively low. On the other hand there is gain at which we aim. Viewing a system represented by a covering provides a continuity of viewing--it's easier to the users to navigate the whole system.
That would be the top subject index file. It presents the top level under T, and four lower levels under A B C D respectively (all four forming the level just under the top level).
There would be also links from A B C D just under T to the lower A B C D.
Next one would need new four different files which present the fragments of the tree under A B C D as top nodes of these more detailed files. I'll present only one below (the rest would look similar):