Tinderbox Forum

Link Actions and Interdocument

I see this thread : Link action examples?

I believe I will use link actions heavily. It looks as if the vocabulary for TBx document to TBx document links is restricted to the three noted in

Is that a hard limit or would cross-document experiments be worth it?

These are on-link actions. Has anyone tried to retrigger them by a stamp or other action?

You might be better asking Eastgate directly about the state of/plans for the TBX URL schema. My notes at aTbRef represent what I’ve gleaned from release notes.

The event that triggers an on-link action is creation of the link and can’t be called via action code. Short of deleting/remaking the link that event won’t fire again. The same code could possibly be run via other action code means but you’d need to use functions like linkedTo() and linkedFrom() to determine source and destination if these are to be used in the desired action.

Why do you want to re-trigger the OnLink? I ask as there may be another way to the same end.

Dunno is the most honest answer. I’ve been with TBx since Web Squirrel and I think Hypertext 96 in DC. As things developed, the element that most attracted me is typed links. These had seemed to me, since the late 80s at DARPA, the one thing that could be disruptive

I invested heavily in TBx because of this, with the notion that I could author in TBx and execute what are now called reactive functions externally. In an early project, I was using the environment to instance a strictly defined set of experimental insights, not taking advantage of two key strengths: spatial thinking and emergent structure. This time I am emphasising these, particularly the just start and discover what and how as I go.

My project is to understand signals in the central nervous system. Turns out, this is a dark art that needs new modelling approaches because many signal types have unknown influence on others, and are largely unknown themselves. It is an amazingly complex systems problem with many fractional insights from others.

A possible approach is to model physical elements using conventional hierarchical note ontologies. This is pretty much done in the field, because we can see what exists. So, notes for things, agents. Then to model influence by links. This has an advantage in modelling the brain as it actually works (so far as we can speculate) and also inform an executable model in some exported artefact.

That puts some heavy semantic and dynamic requirements on typed links. As I say, I’m just a year into this, so just starting and looking at what I can do. A strategy would be to take what are actually named links, and correlate names with ‘effect notes’ that have structure and associated code, so that assigning a link notationally associates that effect.

In any meaningful example (starting with PTSD, but growing to olfactory neural regeneration, and how HSV spoofs the immune system) the model has to be reactive: a change in a note may change notes ‘downstream’ in the influence chain. Notes will change state in a fabric.

The primary novelty we are bringing to this enterprise is that neural propagation is something like narrative dynamics.

So no particular request yet, but I am finding that the files I am creating are easy to break. In my past project, agent/action code dragged things down. So if I can modularise and have links span files, that would be great. That possibility is what prompted the question.

– T

1 Like

I assume the last is reference to your blog TBX, which I well recall. Since then, Eastgate has (unseen) put a lot of work into improving agent performance making such large agent use much better. It is still possible to have too many agents/actions but that judgment is very contextual, IOW one simple can’t say “use no more than N”. Often breakage is fixed by review and incremental formalisation to refactor existing pathways.

The current tinderbox:// pseudo-protocol (here) does not—currently—allow getting/setting values via this method. An email to support (info@eastgate.com) might get you a clearer picture as to the status quo and plans for the feature.

Meanwhile, Tinderbox has acquired support for AppleScript which does support external getting/setting of values. So, unless you’ve a strong aversion to AppleScript (some do!) there is a method available, now. You can call a script using runCommand() with the osascript command line (man page here).

I’d agree about link types. Tinderbox, with its deep hypertext roots has a notion of link types. However, my understanding (from erstwhile convos with @eastgate) is interest in link types piqued in the late 80s, then the early Web took hold but was very underpowered in terms of semantics, and so they rather faded from site. On the Tinderbox side, circa cv4/v5 linking actions arrived that were type ‘aware’ (i.e. could set/filter link types). Link actions, were new to v8.0.0. I’m certainly on board with the notion that link types (semantic metadata) on links is very useful and a good vector for incremental formalisation from human perception of knowledge to digital semantic systems. Not before time, the Semantic Web is a shadow of what was originally posited.

Another angle for mapping dependent relationships that might help is FlyingLogic, a Java based app. It builds off Theory of Constraints concepts but does allow logic based maps. I’ve had the app for a while (because of its map/spatial angle) but never found a real use for it. There is a user forum, very light traffic. The app is kept updated so I assume there are users. The low traffic may relate back to the fact the app seems to have been built of the back of unspecified work done for a lettered agency, so the users may be off the main public web.