The design of the PRD assumes immediate response for every link. If you have to wait after each mouse click, it may be frustrating. In this case, I recommend you to install a local mirror. To modify the structure because of the low speed of the current internet would be a failure - in near future the response will be immediate at least for pure text.
Another decision was not to include pictures. This decision was not only caused by an attempt to avoid problems with child porn laws, but also some kind of protest against the tendency of media to become more "flat" using more pictures and less text. I believe that there are enough people in the world who remain able to read a text even without any pictures.
All PRD-internal links are local links. This allows to move the PRD as a whole from one site to another leaving all links correct and local.
The internal directory structure is not very essential for the user who can simply follow the links. Nonetheless, it may be useful to have an information about some conventions I use - for example if you want to find something having the PRD at home, but maybe to help you with some recommendations if you decide to create something similar for a different problem.
The design is often subject of change. This makes it dangerous to make external links into the PRD. Nonetheless I prefer at the current time to make a change in the directory structure if this leads to an easier and more consistent structure.
There are a lot of internal problems with the directory design. I want to collect similar material in the same subdirectory. But the classification of a piece of information into one of different folders is usually not unique, typically different problems are connected in a single piece.
In the following I list some of some principles and decisions I have used in the order of their success in this application - the PRD.
This is the most clear and unique possible subdivision. For every piece of information the language is well-defined. For every user usually also. Thus, the language subdivision is the most basic and unique.
In the case of the PRD, caused by the languages I'm able to use there are two main languages, English and German. It seems that for every language an own manager is necessary. It is also remarkable that even having the same manager for the English and the German part these parts are far away from beeing identical. This depends already on the different sources available, but also on different sort of discussion typical for each country.
Nonetheless, a design principle was to use the same directory and file names (that means usually English words or abbreviations) for similar things. This allows to automatize a language cross-reference link
The article name convention is the following: Surname of the first author, with a single uppercase letter at the beginning, followed by the year of publication. In the initial phase, I have worried a lot about different rules for the different types of possible necessary exclusions. But until now the number of such exclusions was so small that they don't really count.
It is not easy to have all the index files in each of the subdirectories up to date. Currently, they are not. Nonetheless, it is good to have some files mostly designed to include links to other files.
It is obviously difficult to subdivide all information into different themes, especially because of the cross-relations. But nonetheless this was a successful strategy. The problem of cross-relations may be solved by WWW links. Thus, having a file in the wrong directory is not dramatic.
My first idea was to use type-related subdirectories, that means subdirectories which are based on some common type of data which may be included into these directories. Some of them have been successful, other not. The type-related subdirectories which have survived are the directories where the type was in some sense also related in some sense with the contents: