The outline will be broken into abstract and spatiotemporal divisions: philosophy, math, and much of science (both natural and social), for example, will be in the first division, while history, geography, anthropology, art history, and natural history will be in the second.
The difference between abstract and spatiotemporal divisions of the outline comes down to this: if the point of some chunk of text concerns some event or place, the chunk is in the spatiotemporal division; if none in particular, it is in the abstract division.
The following is a very rough proposal in which I try to model the principles I used in making decisions about how to organize the nodes of my Leviathan outline.
Child nodes should in some way be about the topic of the parent node; e.g., the definition of, a part or aspect of, the explanation of, be a type of, etc.
What the Collation Project has to determine is the extent to which these ordering relationships can be systematized. Generally speaking, they should be placed in order: first one discusses the definition of the item; then its parts or essential aspects; then its explanations (causes); then its types; etc. It remains to be seen what the most elegant canonical list of ordering relationships for text chunks will be.
Nodes that live at the same level of the outline should all be of approximately the same degree of abstraction.
(Of course, how do we determine "same degree of abstraction"?)
For the sake of manageability, it is permissible to promote a header that represents an entire area of study to a top-level node.
Notes: the place of a study should be indicated, if possible, in the outline of previous areas of study (as cross-references). It should be possible in principle to collapse the entire outline so that it has a single root node ("being," perhaps, or "object").
Different nodes that are thus made top-level nodes should be listed according to their relative positioning in a completely "transcluded" (unified) version of the whole outline.
While chunks may live in different places of the outline, nodes and headings may not. If a subject has aspects that fall under multiple headings, the aspects should be placed under those headings and labelled.
For example, the study of religion has philosophical and sociological aspects. While the philosophical aspects might belong under a very broad, general heading such as "Being," the sociological aspects belong under "Society" or some such.
History, geography, anthropology, natural history and aspects of biology, criticism and cultural history, etc., are grouped in roughly the ways they are ordinarily studied.
We recognize the in-principle arbitrariness of temporal and geographical divisions; but we must make some divisions or there is no outline. So the general rule is to use the usual markers for eras and periods, continents, countries, cities, etc.
Top-level nodes of the spatiotemporal outline are eras of history, which are divided into centuries, periods, decades, etc. For each significant temporal division (and there will be a canonical list), there is a fully articulated taxonomy of place (but not identical from division to division); and for each distinct place-at-a-time node (e.g., 18th century France; the WWII years in Germany), there will be sub-nodes created, as necessary, for different aspects (political, social, economic, geological, natural) of the place at that time.
This is just a proposal.
Select a chunk that makes one point, i.e., that can be described as having a single linguistic (pragmatic) function: stating (and elaborating) a proposition; argument; explanation; description; etc.
Justification: more than one point makes a text inherently more difficult to classify, and harder and less interesting to compare to other texts that address the same point. Notes: a canonical list of chunk functions should eventually be created. Also, sometimes Hobbes listed arguments on some small point, tediously, and so in the same outline all of the arguments are grouped together as one long argument. That violates this rule.
Include only whole sentences in a chunk.
Justification: we want the contents of the outline to be easily displayable in other languages if translations of texts are available. Translations can be matched sentence-by-sentence to the original, but not (reliably) in a more fine-grained way than that. Note: the Leviathan outline doesn't follow this rule.
Place no more than ten chunks under a single node: after that, seek to make distinctions.
Justification: one way to view the goal of the Collation Project is to group like texts together. If, in doing this, one ends up with a very large number of texts, one can make distinctions and thereby Note: after very many texts are collated, this rule might have to be set aside in some cases, if there are no further meaningful, useful distinctions to make.
Each chunk belongs under (is described by) its heading.
In other words, do not place a chunk with a set of other chunks merely because it loosely resembles them; it must actually concern the topic of the heading.
One and the same chunk can be given different summaries for different nodes where it is filed. The summary of a chunk should reflect its placement.
For the spatiotemporal outline, a chunk that concerns overlapping periods or places (e.g., Germany and France, 1930-40) should either be placed in the narrowest covering period and place (e.g., Europe in the first half of the twentieth century) or in the period and place that it most centrally concerns (e.g., World War II).
Obviously, the chunk filers should use common sense.
Do not create a new outline node unless (1) there a chunk to place (at least provisionally) under it, or (2) there is a sub-node under the node which is itself created specifically as a place for a chunk.
In other words, the outline is not created a priori. It grows exactly where needed.
Write node headings so that they will fit multiple chunks from multiple sources.
In other words, don't word a heading so that it fits only the chunk in hand. This requires some expertise in the field--it requires remembering (or being able to anticipate) what topics are shared in common across authors.
In first constructing the outline, or in dealing with new problems that arise, edit the outline vigorously.
Check to make sure each child heading belongs under its parent, i.e., follows the node relationship rules.
Virtually all parts of a text are to be chunked--not just the "interesting" parts.
If one changes a node's heading, he must check to make sure that the chunks filed under it are correctly filed, and all sub-nodes so that they are coherent as sub-nodes.
The outline should always be edited so as to be broadly inclusive of all views, not being recognizably in favor of any part of the political spectrum, any particular religion, etc.