I swear, so much complexity of my life right now comes from me wanting to be able to graphically draw out an interconnected hypergraph but also have a convenient textual representation of said hypergraph
I'm sure this makes zero sense to people. But ugh. It's so frustrating to have the ideas in your brain and just not be able to really tease them out in a useful way for others
Signed,
trying to figure out how to map "do the platform engineering thing more better" into strategy and architecture
@hazelweakly ugh yea. Ive explored a few text based ways of generating graphs or similar (mermaid.live, plantUML,tikz...), and they are all kind of terrible. I think part of the problem is that representing the relations between the items is one thing, and saying how you want to display those things is another, and it's apparently really hard to do both well.
@gvwilson@Scmbradley yuuupp. I wouldn't even be mad if I had to draw it by hand (let's be real,I'd prefer it)
but oooff, it would be so nice to just have a vaguely human readable text dump SOMEWHERE so I can search and replace things. That's literally all I want
(Well, copy pasta would be useful too)
(I want like 50 other things but let's start somewhere)
@hazelweakly Maybe abusing stuff like Scrivener? I keep wanting the VR people to stop doing dumb stuff and give me a 3d hypergraph "peg board style" thing instead that just work with hyperlinks and all...
@hazelweakly I relate to this. Text scales, but requires serialising the graph, which doesn't convey the interconnectedness that makes whatever I'm trying to say important or powerful. To be comprehensible to other people basically means breadth-first traversal, and now I'm explaining a whole tech tree of concepts, and most of my readers have lost interest.
And then you update the graph and you have to redo the traversal!
Sometimes I just want to visually ontologize GADT type signatures, TLA+, nCat, or traits/interfaces/header files. Bonus if it spits out graphviz.
Sometimes I just want a hierarchical state machine or dag that hides detail.
Sometimes I dream about building a graph editor in the browser, based on a very particular set of solutions ( #veilid for p2p, #egui for editor, #bevy for data, and #rapier for deterministic physics based layout). Layout is the really tricky part for building a graph editor I've found.
@hazelweakly that feeling when my mental model of the problem domain is this seven dimensional object, and i’m forced to describe it with cross sections to the other squares in my 2d plane of existence, and they believe they can gotcha me because they can’t tell the difference between orthogonal projections and inconsistencies in reasoning
@hazelweakly if you see a hyperconnected graph, that's a sign that there are missing dimensions to the domain that could simplify things a lot: oblique strategy - normalize the data model. I try to find the things that cast shadows and to give them names.
By analogy: imagine a bicycle manufacturer that does not have any understanding of the term "bicycle”. They put together bicycle parts in a consistent way, but when it come time to ship them to the store, they take the bike apart and put each part in a separate envelope. Then they leave it to the store to put the parts back together solely by the order number. God help them if someone orders two bikes at once.
@flyingsaceur I can normalize the data model, it just drives me up the wall to try and reason about 25 different things connected in 8 different ways with a whole dependency graph and weighted adjacency matrix going on...
Normalizing that is totally doable but it'd involve making like 180 different pairwise connections and referencing things by increasingly baroque IDs and essentially building a highly relational algebra out of tables when all I wanna do is draw cutsie lil arrows n circles :'(
create a dir of .html files, each one is singular resource or concept or whatever.
write your plain text description of what the thing is. eg: /sam-vimes.html => Sam Vimes is the Captain of the City Watch of Ankh-Morpork
Create <a> links between the resources, like: City Watch of <a href=/ankh-morport.html>Ankh Morpork. So far just making a hypertext website. Easy!
@hazelweakly 4. To the <a>, add a property which describes the type of relationship – bonus points of its a uri but it can be whatever: <a property="ex:livesIn" href="/ankh-morpork.html">Ankh Morpork</a>
5. Write out all the things and their relationships like this. You have a hypertext document with plain text and dense links.
6. This is RDFa microformat data! Scrape it out of the HTML and convert it to JSON-LD – with something like this! https://gist.github.com/nikolaswise/d500a4bf7d76134274377b8b47226301
2/??
@hazelweakly I've been using this for writing documentation, enriching text with autolinks, keeping text strings fresh across docs, and managing redirects. Not a ton of visualizing yet, but this has really motivated me to get some sample code up and write a longer explainer of how to use this for good.
It reminds me a lot of the markdown zettelkasten stuff people are doing with the visualizations they're building on top of that. But the jsonld approach is new to me! Thanks for sharing! ❤️
@hazelweakly Maybe onlinewardleymaps.com could help? It’s got a text definition that can be versioned and saved in a repo and a strategy graph structure.
@adrianco I think the big thing I'm struggling with here is that the most useful artifact of that style of diagram is the conversation that happens when building it. But I had the conversation with a bunch of different people over time in different ways and now I need to reconcile it all together in a way that brings them all along on it
Which means I can't really just draw out the map. (And even then, there's buckets of text per node here, so it feels odd)
In complex landscapes, a complex map is going to be of little help to people who aren't experts at reading the map.
Maps are models, so it may be worth making multiple maps for multiple purposes and dedicated lenses/perspectives, usable in distinct contexts.
(analogy: aeronautical charts, topographic maps, demographic maps, and road maps may all be connected but are distinct as contextual usage outperforms completeness)
@mononcqc@adrianco yeah I think it's very much that. The making multiple models is the data normalization thing I run into, I think. I end up doing this anyway just because it makes sense but it's nice to have "better" ways of making multiple models or tying them together so that I can evolve them a little Better and so on
(Ideally "lenses" would let me have central data and create the updated view better. But I don't really have a nice way to do this semi adhoc)
@billseitz@mononcqc@adrianco mostly 1, with a little bit of 2 and 3. I've managed to get unstuck and figure out what I need to do, it's just something that requires talking about it to quite a few people and being mad at how hard it is and then getting some advice and noodling on it until something clicks :)
@hazelweakly@adrianco The graph + text + collaboration + multiple views and projections for understanding -- this process sounds like an advanced use of Federated Wiki.
One community member, Thompson Morrison, has used FW for collaborative authoring. Refining nodes and connections through conversations among the authors is part of the process of creating a navigable knowledge base which can be viewed in multiple ways and also be extracted as a book or books.
“ I think the big thing I'm struggling with here is that the most useful artifact of that style of diagram is the conversation that happens when building it.”
Welcome to the world of designer, visual asse(r)ts of Information Architecture (IA).
Hypergraph (capture and sense model) for audience of 1 perspective - you, needs to be translated and “transpired” into conceptual models for audience of many and different perspectives.
@dahukanna@adrianco that's a biiiig mood and it makes so much sense. I just wish we had better ways of making those documents in a way that we could translate them better, y'know?
@hazelweakly@dahukanna Have you read Dave Anderson’s Value Flywheel Effect book? There’s examples of several different variants of Wardley Maps and other tools being used. I’m biased (wrote a foreword for it) but I like the methodology.
@v1sea it's a hypergraph because most of the nodes are themselves graphs which have edges that span across to other nodes in the outer graph(s). And it can keep going N layers deep. Because of course.
@v1sea I don't, but it's worth noting that most complex multi dimensional relationships can be modeled with one, and in particular, "knowledge representation" as a more abstract concept
But even further than that, basically any multi criteria weighted decision making algorithm thing for group consensus is going to look like a hypergraph when you're done. Even if it never gets materialized
Like the far extreme antipode of Engelbart taping a brick to a pencil: sometimes my mind is just certain there has to be a better, more expressive tool to use for reasoning
Add comment