Here are some extemporaneous thoughts about dialogic publishing, loosely in response to http://wordpress.liquid.info/10/pdf-notifications/frode/.
No authoring system will satisfy everyone, but a common structure of dialogue (truly, a common semiotic structure) is, in my belief, while not perfectable, pragmatically feasible. Such a universal system must be based on an interoperable format/protocol. I will refrain from going into technical details at present. I just wish to frame/name this core part of the problem.
I think the social and group dynamics of the intended interaction are important. Is it 1-on-1 dialog? 3-person? More? Each of these forms has different dynamics, just as they are in-person. For example, when a third enters a 1-on-1 conversation, body positions and eye contact shift. Current digital dialog/collaboration platforms have largely neglected this aspect.
I believe that quality of discourse is mostly independent from length of content, in the sense that a quality conversation can (potentially) be had at any length of retort. The number of participants is a dependency however. I have observed empirically in psychodramatic group processes (free association groups) of up to 200 people, that as you increase the number of participants, retort size goes down (more slices of the pie = smaller slices) and become punchier. Quick tropes such as humor and more emotionally evocative comments are leaned on to make up for the shortened duration. These dynamics should naturally be considered in digital spaces as well.
Dialog (of any number of participants) is at its best a consensual act(ivity). The dynamics of choice, consent, election, especially considered diachronically (as a changing, ephemeral, negotiated status), adds considerable complexity to the situation.
At its heart, dialog depends on relationship. Thus I advocate for a relationship-oriented approach to digital dialog.
Publishing is an important part of discourse, though mass publishing as conceived often ends up being more of a weapon than a tool. I would be curious about a system that reframes publishing (of dialog) as exposing or making transparent. Thus the work is referencable/addressable without removing it from its primary dialogic (relational) context.
One can help orient the user to the territory of the dialog by providing a (ideally interactive) diagrammatic topology of the dialog. This remain simple in the linear 1-on-1 dialog form, but quickly gains complexity when additional participants (either officially invited/sanctioned or otherwise) are added, or when the authors begin playing with linearity and time, such as exemplified in Katherine Hayles’ early experiments in hypertext dialog.