I think that it will be easiest to understand xanalogcial "overlay" links by contrasting them with the embedded-anchor "jump" links that we're all used to.

Embedded Jump Linking

In systems such as the Web, links that users can see and interact with[1] are implemented by embedding anchor tokens into the text stream of a document. An anchor token may be empty (and invisible) and serve as a named destination point for a "jump", or they may contain some kind of visible text/imagery that the user interacts with to trigger a jump [what I mean by jump is clarified below under the Presentation heading].

These embedded links are a part of the document's structure, just as inline citations and symbols indicating footnotes and endnotes become part of the structure of a printed document. Because links are integrated into the document structure, in order to see additional or alternative links in the same text, another document must be prepared. In order for a user to add links to a document, they must either create their own copy of the document for editing, or they must be the owner of the original document and have some means of editing it and saving its changes on the server.

Embedded links typically cannot overlap. In an early, pre-Alph system I worked around this by using named start/end tags, but this is not how links are implemented in Web browsers and the majority of other embedded-anchor systems.

These links are one-way; that is, destination documents do not provide links back to the documents that link to them. This, too, has been addressed by a small number of "linkback" schemes on the Web, which tend to work by keeping a database of referrer documents on the server, then dynamically generating return links at the foot of a document when it is served.

Presentation:

Embedded jump links are typically indicated by colored, underlined, or otherwise decorated text, although the contents of a link anchor – the bit that the user interacts with – can contain any portion of a multimedia document. The way that links are decorated can be chosen/designed by a document's author.

Traditionally, these types of hypertexts have been navigated by "jumping" from one document to another: the user activates a link, by clicking, tapping, or selecting with their keyboard, and a new document replaces the one currently on-screen. A history of visited pages is kept that the user can move backwards and forwards through, although it's worth noting that this history is non-branching. For example, if a user views document A, jumps to document B, then to document C, jumps backwards two steps to document A again, then jumps to document D from A, the user can no longer jump "forward" to documents B or C as they have been pruned from the history. Contemporary browsers provide tabs, or overview pages with document thumbnails for moving "laterally" through documents, though each of these tabs has its own history, with no record of the document that the new history was forked from. The ability to open links in a new "popup" window was abused by advertisers in the first decade of the Web, and since then most browsers will instead open new documents in tabs (or in-place), and the assumed mode of presentation for each document is full-screen.

Overlay Linking

In a xanalogical system, links exist independently of documents. A document may bundle a set of links (a graph) with it, but the links are not integrated into the document's structure.

The endpoints of links are not anchors embedded within a document (or necessarily entire documents/resources), but instead are pointers – references to specific fragments of primedia. As xanalogical documents are transclusion-based, this effectively means that links point "through" the presentation layer (the document) at the media underneath.

Links, being independent of documents, can remain "open" in the browser even after a specific document or resource has been "closed", or removed from view. Any document subsequently loaded into the workspace that transcludes media referred-to by an open link will have link highlights or symbols overlaid onto it.

The user can easily create links to and from any document or media resource, as they do not need to create editable copies of documents or resources in order to embed anchors into them. These links can be published and/or privately shared with other users.

Multiple sets of links (graphs) can be active in a workspace at one time, and the user may enable/disable the display of individual graphs or link types as needed.

Links can overlap each other.

Because the user's browser is aware of both endpoints of a link, they become two-way. When one link endpoint is removed from the workspace, the browser will continue to indicate from the other endpoint that a link exists to be shown.

Links stay connected to content, even when the presentation changes. If the user links a note to a paragraph in document X, and document X is subsequently revised in such a way that the linked paragraph is removed, the link remains valid as long as the primedia has not been tampered-with or removed from the network.

This also means that users can easily see when the target of a link is being presented differently from how it was presented when originally linked to.

[Show an example of how a paragraph that has been linked-to will have a different link indicator if it is revised by removal of a phrase, or the insertion of a new one.]

Presentation:

A key xanalogical system principle is that the user must be able to see side-by-side documents with graphical indicators (lines, arrows, highlight "bridges") that point across documents at the the endpoints of links. "Parallel pages, visibly connected"[2] is a core paradigm. This isn't simply an aesthetic conceit: studies have shown that it is beneficial to users of hypertext systems to have access to a graphical overview or map that shows the connections or groupings between hypertext nodes[3]. Because a xanalogical system should support the creation and dynamic manipulation of hypertexts, having visible links between nodes makes the system self-mapping.

Links may be indicated at endpoints by symbols/icons, underlines, highlights, outline boxes – whatever the user's system has been configured to support. The lines or bridges between link terminals (the TRANSPOINTERS) may also have varied appearances, based on the nature of the link (its relation type, its completeness, ...) as well as user preference.

Linking by reference combined with a transclusion-based document model goes a long way towards solving the PROBLEM OF ANNOTATION.