Roam Research

micahredding

I think we should have a discussion of the approach offered by Roam Research. On the one hand, it's a tool for Zettelkasten-style work. On the other hand, it brings its own conventions to the table—which may provide useful reflection on what we're doing here.

The basic approach, as I understand it so far:

1. A journal, containing a freeform nested outline of ideas.
2. Each line of the outline can link to various Concept Pages.
3. Each Concept Page has a collected search of all the statements made about it.
4. Each Concept Page can have summary-type statements.

This is more complex to describe than it is to use. Some things I like:

1. Lines as the atomic units of meaning. This is mostly how I use my note system now, but Roam makes it the explicit basis of the interface. More significantly, the system itself searches at the line level, instead of just the note level.
2. Using [[double-bracket]] search as linking. Essentially, every phrase in double-brackets is a tag. To click on any double-bracket phrase is to see:
   1. The structure note for that phrase
   2. All the other bracketed uses of that phrase
   3. All the other unbracketed uses of that phrase.
3. Concept Pages collect all of these together. For example, on the concept page for the "Monad", it would show any summary statements, any statements linking to the "Monad" concept, and any statements using the term "Monad" informally.
4. Because these are collected together—and yet clearly distinguished—this adds a lot of value to Concept Pages over regular search.

Thoughts about Bracketed Search:

Of particular concern to me is data portability. Currently, when you export from Roam, it gives you a Markdown-style outline, with double-brackets intact. This works in a general way, but once outside of Roam, this becomes fragile to future changes in verbiage.

This makes me wonder if we could work towards a more general "industry-wide standard" for Zettel links. The general practice here (and what I do) seems to be linking notes like this:

[[201912032209]] The Monad

While Roam exports them like this:
[[The Monad]]

It would be nice to allow some kind of both-and, such as:
[[201912032209 The Monad]] or
[[The Monad 201912032209]]

But this currently seems inadvisable in The Archive, as that performs an AND search, rather than an OR or fallback search.

Ideally, it seems like you should use a format like [[The Monad 201912032209]], but the search should work something like [[201912032209 XOR (title: "The Monad")]].

orderinthecourts

@micahredding said:
It would be nice to allow some kind of both-and, such as:
[[201912032209 The Monad]] or
[[The Monad 201912032209]]

I was thinking of suggesting a new feature to The Archive to allow titled wikilinks in the form of [[201912032209|The Monad]] as per the MediaWiki specs, but Christian and Sascha seem to be too busy at the moment. Hopefully once the course is done and app development resumes it can be added to the queue.

ctietze

Please do not refrain from requesting anything. I can put this into my abysmally long queue for the time being Maybe it makes sense to separate single requests from this reflection in their own discussions. Because atomicity and all, and to keep this space for meta-discussions.

micahredding

I second the MediaWiki-style links—that hadn't occurred to me, and I think it's a cleaner and more semantic format than Markdown-style links.

I'm curious. How are people using (The Archive's) bracketed search in general? The current way the search works allows for full boolean searches to be included in notes themselves. This is very powerful, but does not seem to optimize for the most common use case, which is a search for a specific phrase.

For example, clicking on [[The First of December]] in The Archive would locate notes including December The First of, The of December First, First December of the Year, and of course, any note which included any of those words inline in any order. It would not, however, locate notes which included only First of December.

On the other hand, I could use The Archive's built-in functionality to make it work the way I wanted it to. For example, [['First of December']] or [['The First of December' OR 'December First' OR 'First of December']] or even [['First of December' OR (!January AND !February AND !March)]].

As a software developer, I kind of love this. But I'm not sure of the practical value. In general, I find myself using this kind of complex syntax to approximate a behavior I consider more in line with a literal match.

My suspicion is that most of the people here really only use brackets for [[201912041103]] zettel ID-style links, and so this doesn't come up. However, I still have an intuitive feeling that bracketed terms with no IDs should work more like tags—aiming for something closer to a literal match. Or perhaps even a hyper-literal match: [[First of December]] could match solely with other occurrences of [[First of December]].