Outliners, trees and meshes

Thanks to Julian Bond for his response to last Thursday’s Outliners post, which pointed me to his essay from 2004 on outlining, “Outliners Considered Harmful.” (The title, I should point out for those of you who are not steeped in programming lore, is a nod to a tradition in that literature dating back to software pioneer Edsger Dijsktra’s 1968 paper “Goto Statement Considered Harmful.”)

When you use a tool that encourages you to think in terms of hierarchies, everything looks like a hierarchy. Unfortunately the world is much much messier than that. Almost everything is actually a mesh not a hierarchy. And when hierachies do exist in the data, it’s very likely that you will find 2 or more inherent hierarchies that are orthogonal and in most real world situations it’s more like 10…. I suspect that the move towards anarchic tagging in systems like del.cio.us and Flickr are driven by this tension between mesh and hierarchy as well though I haven’t seen it expressed like this. Tag driven systems look horribly uncontrolled to hierarchy people. But they actually reflect the real world much better than hierarchies….So in a nutshell, Outliners are harmful because they lead to hierarchy thinking. And hierarchy thinking is harmful because it leads to political hierarchies. And all of this is harmful because the world is actually mess(h)y and not structured into elegant trees.

I’m reasonably familiar with this argument. For one thing, it’s part of a discussion I closely followed while researching my book at OSAF. During some of the early work on Chandler, its designer, Mimi Yin, inspired by Christopher Alexander’s celebrated essay “A City Is Not A Tree,” tried to figure out a structure for personal-information management software that privileged “semi-lattices” (what Bond calls “meshes”) over hierarchies and trees. (Clay Shirky’s 2004 post on the Alexander essay is worth rereading.)

I’m also familiar with it because it is a theme that David Weinberger has been pursuing over the last several years — a pursuit that is culminating in the publication next year of his book, Everything is Miscellaneous, which, he tells us, he has recently completed, and which I’m eagerly awaiting.

The thing is, I’m not actually a particularly hierarchical thinker. My love of outlining is less a matter of obsession with rank and structure than an appreciation of flexibility. I don’t especially care that outlines are built around parent and child nodes neatly arrayed in tree structures; what matters is that outlines give me easy handles to move chunks of loosely structured information around, and they let me quickly zoom from a low-altitude view to a high-altitude overview and back.

All of which may explain why I still love Ecco Pro so much. For one thing, Ecco doesn’t force you to follow the outline structure rigidly; you can drag nodes pretty much anywhere you want. Even more important, each Ecco outline can also be fitted with “columns” — really, versatile tags or categories (they can be free text or checkbox or dropdown or date) that provide exactly the sort of cross-meshing or semi-latticing that Bond rightly reminds us we need. You get the best of two worlds — outlining structure and free-form mesh-i-ness. And it’s very easy to adapt to the David Allen “Getting Things Done” method for those who’ve been bitten by that bug.

I’ll stop raving now: Ecco is old, unsupported and probably has no future. Still, it’s rock-solid, and free, and it continues to serve me better than anything else I’ve found. As for the “Outliners make your brain too hierarchical” line, it might hold for some of the simpler outlining tools out there, but I really don’t think it applies to a program this versatile and fluid.
[tags]outliners, pims[/tags]

Post Revisions:

There are no revisions for this post.

Get Scott’s weekly Wordyard email


  1. Me too. The only thing that approaches Ecco for flexibility is MS’s onenote, which I find in some ways too flexible. I don’t like its close binding to Outlook, either, simply becasue I don’t use Outlook.

    Perhaps this is because I have internalised Ecco’s grids – they were such a wonderful UI idea – but I find that taggig things in columns as I can with Ecco is quite helpfully rigid conceptually, while being fluid on the screen.

  2. Hey Scott,

    I tried using Ecco Pro — it was a little rough for me, but I suspect my mental map doesn’t work well with how it’s laid out. But I’m oddly fascinated by the application, nonetheless.

    Not that I have time to pursue such a thing, but do you think a Open Source, Reverse Eng. version of Ecco Pro is doable? I’m slightly surprised such a beastie doesn’t exist, yet, given the community surrounding it…

  3. Well, there’s always BrainStorm (for Windows) which was written in 1981, a long time before Dave Winer’s ThinkTank and actually published a couple of months before, in November 1983.

    When I wrote it, I was ignorant of outliners, Engelbart and Winer. I knew about Tony Buzan’s mind-mapping and various theories of how we think. I just tried to replicate this in a textual computer (CP/M). It’s still textual and it still sells. It has avid fans.

    You can think and work how you please – non-hierarchical, hierarchical, a mix of both. It has namesakes – stick something in and if it’s already there, anywhere in the model, it shares the same descendants. (Or not, you can switch it off when it doesn’t suit.) You can change your mind and, after learning half a dozen pretty obvious things, the program doesn’t get in the way of your thinking.

    I showed BrainStorm to Dave Winer at the 1984 Softcon show in New Orleans and, towards the end of an interesting conversation, said ‘this is what’s missing’ from ThinkTank. It wasn’t long before ‘clones’ appeared and I kicked myself. They’re not the same, but I would say that wouldn’t I?

    Apart from the namesake stuff, BrainStorm is not matrix-like but, like David Weinberger advocates, it can hold the same thing in multiple branches of the structure. (I know his views relate to taxonomies, but it applies to thought as well. He knows me as ‘the English guy with two business cards’.)

    Oddly, we (Caxton Software at the time) chose ‘ideas processor’ as our slogan and Dave Winer chose ‘idea processor’. We came at our discoveries/inventions from different directions. I was coming from the ‘how we think’ direction and Dave was coming from the ‘how we work’ direction. At the time he was programming in Pascal and he’d been watching his wife (I think) working on outlines for her university work.

    After years of paddling our own canoes, I notice that OPML is joining our worlds together. We are working on native OPML output from BrainStorm. I am a Grazr fan and have been pumping BrainStorm models into the OPML Editor then, from there, onto a web server from which they can be rendered by Grazr.

  4. I’m a huge fan of outliners, but generally agree with Julian.

    If you step back a little, there are (at least) two aspects to outliners: the hierarchy as *view* of information, and the hierarchy as a *model* of the data. In short I’d say the former is great, the latter broken when applied to real-world information (and the Web).

    As an example of how to get the best of both worlds, check Tim Berners-Lee’s Tabulator. It allows you to browse data on the web in an outline. However, the underlying data isn’t limited to being hierarchical – the model is the Resource Description Framework, which can be seen as a node and arc graph model of information.

  5. Always good to see a deliberately provocative piece generate a little discussion. grin.

    Having poured scorn on outliners, I’ll completely reverse the position and say that I love them for brainstorming when the final output is going to be linear text. As Dave Winer says (if I read him right) in the blog that points here, the ability to constantly edit the hierarchy is a great aid to thinking and structuring what ends up as a document.

    My rant is really about what Danny describes above as the hierarchical model of data. It’s one of the things that bothers me about OPML, XML and Outliners used as a way of structuring data. It feels like a straitjacket to me.

    Next week “Tags considered harmful” and the week after “RDF considered harmful”!

  6. I’ve always been a big fan of outliners, mind mappers, and thought processors of various forms. Lately I’ve been alternating between Dave Winer’s OPML Editor [1], Eastgate’s Tinderbox [2], and various wikis [3] and half-baked things of my own [4].

    As for the perils of being locked into hierarchy – I can say that I agree with the sentiment somewhat: Thoughts in my head don’t usually settle well into parent/child or container/contained relationships – or even a linear list order. Jotting them down in such a structure can prematurely commit me to the structure, no matter how flexibly I can juggle the structure around. Once I’ve gotten notions out of my head and in front of my eyes, that’s the form that starts to gel.

    The one tool I’ve used that seems to help preserve vagueness for as long as possible is Tinderbox, with its freeform 2D views which slide easily into outlines and other forms. With a 2D map, I can just quickly toss notions onto the screen and sorta-kinda place them in fussy groupings without deciding order or precedence. Over time, I can sort the ideas on the screen, and eventually maybe commit to a hierarchy.

    Picture a pile of index cards, a marker, and a big coffee table. Jot down ideas and fling them onto the table in haphazard piles. Once you’ve run out of mental output for awhile, then you can start arranging. Tinderbox gives you that, albeit with a bit of a learning curve to exploit all its features.

    None of this is a real criticism of outliners in general, though, since eventually I end up back in outlines. I just like to maintain a lot more indecisiveness up front while I’m emptying my head.

    [1]: http://support.opml.org
    [2]: http://www.eastgate.com/Tinderbox
    [3]: http://flyingmeat.com/voodoopad/
    [4]: http://decafbad.com/trac/wiki/XoxoOutliner

  7. # Asther Findley Says:
    “Broken is pretty serious, how do you know is broke?”

    Try putting your family tree in a tree – maybe start with ancestors from from yourself, then try descendants from one of those ancestors.

    # Julian – yep. I think Dave is attempting to dodge the issue, an editable hierarchy is still a hierarchy, he’s invested too much ego to admit any flaws.

    # Les – I’ve tried a lot of outliners/mindmap tools, none of which I’ve found entirely brainstorm-ready. Dave’s editor I find too dependent on Dave’s Own Stack, I don’t want to work that way. I’ve still not tried Tinderbox, that’s me being a cheapskate.

    Which reminded me I’ve not looked at Mindraider [1] for a while, and while searching (couldn’t remember its name) I ran into an outliner for Moz [2], but I need to restart FireFox before I can comment on that…

    [1] http://mindraider.sourceforge.net
    [2] http://outliner.mozdev.org/

  8. Brett Glass

    I worked with Dave Winer in the early days of Living Videotext, which published his ThinkTank outliner. I took the job working on it specifically because I liked the idea of being able to use the outliner to build, edit, and transform hierarchies AS hierarchies, retaining the substructure as I moved parts around. I still use outliners to this day to organize my thoughts and my schedules.

    Yes, it’s true: Outlines — even outlines with “clones” and “links” — aren’t always sufficient to edit a more “messy” structure and don’t do a good job of representing a table or grid. They’re sort of “one-and-a-half-dimensional,” neither one-dimensional linear lists nor fully two-dimensional grids. But they’re good when you need to flatten information into the “almost linear” format you need to present it as a presentation, an article, an essay, a report, or a book (which are, when you get down to it, linear flattenings — or depth-first traversals — of an outline).

    Jerry Seagraves (the author of Borland’s “Lightning” instant spell checker) wrote a program which expanded upon the concept of an outliner by allowing you to “invert” outlines, exploiting the relationships between an item and its “children” and “parents.” It was incredibly clever — so clever, in fact, that most people to whom he showed it just did not understand it. It never became a commercial product.

    Ted Nelson’s “ZigZag” (see http://xanadu.com/zigzag/) is even more powerful in that it has multi-dimensional tables, random links (no surprise, coming as it does from the “father” of hypertext) and any number of “views” of the data. But again, it’s so far above the way most people think about thinking that it leaves most people scratching their heads and then going back either to linear lists, outlines, or two-dimensional tables.

    So, I wouldn’t say that outlines are harmful. They’re a step up from linear lists (which they subsume) and offer ways to show and hide detail (which two dimensional tables can’t do) and create a structure that’s flattenable for presentation in serial form. That’s what they’re good for, and that’s what I use them for.

  9. I’m another Ecco Pro user who never found a good alternative. I’m using a Windows/Palm outliner called Bonsai now, which works but lacks Ecco’s columns. (I abandoned Ecco Pro when it became unsupported, but I’m still tempted to go back to it.) Like someone else said, OneNote was too freeform for me.

    I wish 37s would add at least rudimentary outlining to Backpack or TadaList. I’d switch to that in a second, especially if it output OPML so I could switch away when I wanted to.

  10. Will Highfield

    Ecco Pro is my go to program for almost everything I do. Support? Who needs support? It does book keeping, lists, complicated estimating, simple gantt charts for project scheduling, great phone book and calendar functions, and much more.

    But I do have a question. Does Ecco Pro work on Windows Vista? Sooner or later I’m going to be forced to upgrade.

  11. john

    Did you ever find out whether Ecco works with Vista? Have used Ecco since the 90’s, and before that Lotus Agenda. As you say, will have to upgrade eventually….. but I keep holding back as nobody seems to know if Vista will work ok with it?


  12. Dale

    It’s funny, I don’t think the same way. That is, about trees and hiercharchies…

    I love to work with tree-structured information. As the tree matures with it’s useful branches, I tend to grasp a clear vision of my info-base. I don’t really “see it” as a hiercharchy but rather view it as “relationships” linking data together. It effectively “bonds” information and this in turn creates “highlights” for data storage.

    It’s not because you visually see a tree in front of you that you see hiercharhies… it’s more, how you perceive that stored data. For me, that’s not the purpose behind the much-loved tree type storage option. The tree usefully outlines relationships and data links. It’s a visual aid for locating data, that’s what the tool was designed for – bringing information within reach – helping to highlight focus. It basically has nothing to do with world hiercharchies and systems – and not at all. If you’re looking at tree-structured information, and viewing it as such, then you’ve missed the point. (that’s only my opinion, ofcourse)

    But, as a tree’s branches mature, then info storage relationships also mature in representation – if you develop your tree structure as a mature process of data-handling… conveniantly afterwards, well located ideas on the branches begin to develop, and then, so does the related branch. As your ideas mature and develop on the branch, then new ideas are stimulated or generated – just like new branches on the tree will grow – a simple natural progression in life…

    When I look at trees in the forest, I never think of hiercharchies. I tend to think that your interpretation is somewhat harmful and not the outlining process which is a valuable aid for data management.

    …just my point of view…


Post a comment