Lambda the Ultimate

My problem with RDF is that existing applications expect me to do all the work. If I was organised enough to neatly annotate all my information, I wouldn't need an RDF application to assist me in managing it! Google has shown that automated techniques can do a good enough job of relating information, and in my humble opinion automated techniques are the way forward.

I must be the only person in the world for whom a scheme of randomly splayed little blurb boxes simply doesn't work, let alone represent a pinnacle of knowledge representation. This one uses the venerable touchgraph, which has the added bonus of making your blurbs jiggle around like so much jello. What I really want is something like a spreadsheet -- a two-dimensional editing space, but without the silly editing restrictions that typically go with spreadsheets, along with the ability to box and link regions, annotate the links (ala rdf) and so forth. To revisit a familiar refrain, something where the content takes precedence over its presentation.

Mark, I agree entirely that it's time for cool software, and personally think RDF (plus OWL) is the fastest route on the web. "...needs less abstract public relations" - yep, that sounds about right.

Regarding Noel's point, annotation is only one aspect of data. Even there it's not always necessary to do more work - the (XHTML) microformats initiative is demonstrating that it's possible to capture the kinds of data people are already producing in a more machine-friendly fashion (and microformat data can be interpreted as RDF via GRDDL).

Right now there are *huge* amounts of data in SQL RDBMSs that is generally inaccessible and uninteroperable. The same data expressed in RDF/OWL can be shared web-wide. The same goes for all the stuff in the filesystem. Naturally there'll be work to do on access control, but the same applies if you're only looking at text search.

Chuck, to quite an extent I agree (in no small part due to what's in the next paragraph), the blurbs aren't the answer to every GUI need for graph- (and other-) shaped data. The spreadsheet has been cited quite a lot recently as the kind of usability goal for a Semantic Web editor/browser (i.e. not necessarily a spreadsheet per se, but something as good). Tim Berners-Lee was recently playing with a similar sort of browsing idea, see Tabulator.

Funnily enough, I spent rather a long time on an RDF-based MindMapper kind of tool myself, IdeaGraph. It's been on pause for a couple of years now, but when I get some spare time I plan to get back onto it (first task is opening the source), but this time I intend to make it more project-oriented, looking at the general mindmap idea made it really hard to keep focussed, far too easy to play...

See also: Semantic Web Starting Points

Noel's and Chuck's gripes only underscore the point already made. There's poor software selection. Fine, you don't like Mind Raider - be constructive then, suggest what interface you would prefer. Mind Raider was offered as a conversation piece. Whether it "makes you do all the work" is something I neither accept nor refute, since I haven't used it long (nor have you - it's alpha software). Who knows what mix of tables, spreadsheets, hypercards, graphs, or other gizmos is ideal. Constructive remarks or links I'd be glad to see.

DabbleDB has a reasonably good UI philosophy but a poor back end design. Like Chandler it neglects RDF, while everything lives in RAM and the http is insecure. Too many projects neglect encryption issues. They're both cool projects which should exploit RDF, not reinvent it. The interesting SIMILE project reviews some user scenarios to help motivate concepts.

Let's not equate lack of applications with flaws in RDF. Regard RDF as a back end model, not a user interface or view. You need something like RDF to do the job at all. Chandler and DabbleDB are reinventing RDF from first principles. Reading Chandler XML specs about "kinds" and "items" gives me headaches. I would rather use meta-stuff with external interest and support. Maybe Haystack (now based on Eclipse infrastructure) is closer to the mark. I don't want to talk Chandler or DabbleDB down, it's just that clean-rooming open standards like RDF seems silly.

Standards count because of what critics like Noel miss. That is the notion of sharing and hacking ontologies as we do code. The DabbleDB video says just that: (a) beta users already want to publish their schemas and (b) an end user Dabble "ecosystem" is the ideal outcome. That's the whole point of RDF, so why not use it. If a user starts a new hobby, he'll Google ontologies and import his selections into the software. It's the same way with document templates. Find one reasonably close to the need and modify it. Stock forms are sold in office supply stores. Likewise we can pull stock ontologies off the web and modify to suit.

My quibble with RDF's semantic-web big-think is that it robs equal time from personal data (as in "personal computer"). This imbalance may explain why PIMs don't bother with RDF. The marketing is all wrong. Detach the network cable and RDF still works. A private RDF database can use a public ontology, for example.

Some sharing notions don't fly. Few companies use LDAP or calendars to share in any serious way. Those I've seen shared scarcely extend beyond the corporate staff lists and holiday parties. That's why I dislike this emphasis. Contact lists are rather personal and people keep them private. They'll share specific entries upon request. That is the only "access policy" anyone wants to bother configuring, i.e., none at all. So it should go with RDF. Some data goes public, some stays private. We should not prefer one data class over the other. If we only talk about the "semantic web" then we neglect personal data which RDF can handle quite readily. One can obviously generalize individual data to small business data, hobbyist clubs, families, etc.

E-mail is a PIM mis-focus. (DabbleDB avoids this mistake.) The problem of e-mail is big enough to divorce from PIM apps. Almost all PIM projects revolve around e-mail. But many dedicated teams already offer superb clients. Why burden PIMs with that effort? What you want is an interface to e-mail clients, not duplication. There are many neglected categories of personal data out there. That's where PIMs should focus.

And we're not adding user work. RDF can reduce effort required to keep ahead of yourself. That's my whole interest. RDF can defer necessary organization and bypass the unnecessary. At time X it can be hard to know which is which. At time X++ one has a clearer idea.

Think about paper. The paperless office never happened, and, as a grand design, never will. We can still realize small improvements. I like paper. I just don't like lots of disorganized paper in piles everywhere. RDF to the rescue? Cheap digital cameras, scanners, and printers are ubiquitous. Web-cams sit atop monitors; minicams live in cell phones; you can buy scanners for $75. Imagine scanning stray notes into RDF databases. Then toss the paper without a second thought. No more filing cabinets! The RDF notes live in a twilight zone between trash and non-trash. The notes are not "filed" until you annotate and classify them in some way. But nothing's lost if you don't. You're no worse off than with disorganized physical notes. In fact (a) your desk is clean, (b) the notes are primed for classification, (c) even without that, they can be queried by date, and (d) OCR may produce searchable text. The only up-front work was to snap a photo - no ontology, no nuthin'. Annotate later with time, need, and interest.

The software can just date stamp the entry and that's that. All you need to remember later is "oh yeah, it was written somewhere." You can query all notes taken within a certain date range. Later annotations won't face the rigid formats of RDBMS. Ultimately you just use various RDF queries to track and relate concepts. Your ontologies can morph over time as you see fit.

Consider more formal research and writing. The usual procedure is to write notes in a library and then classify and outline in preparation for writing. This is where RDF can help. I cannot take research notes on a computer. It's hard enough keeping three books open on my desk. Nonetheless, actual writing does happen on the computer. It would be lovely for RDF to organize thoughts using written notes as inputs. This application is just one of many possibilities.

P.S. A search engine that does not store all your queries to the end of time, and also clusters search results in a way vaguely approximating these RDF ideas, is http://www.clusty.com. There are Firefox plugins.

I want to sideline resistance factors like scary-new-stuff and make-me-do-more-work.

Developers already perform ontology development for every app they write. Ontologies are not new in software. We just use different lingo like "class hierarchy." (OK they're not 1:1 but still.) See Haystack's design philosophy about breaking down applications. The issue is opening up the internal data model so other apps can use it. Every app has data models and that is really the first thing a developer develops. RDF is just a vehicle for sharing them. So ontologies are not scary and we're already doing them. If we want content taking precedence, per Chuck, then we need to promote these models above associated application views. RDF is one means of doing just that.

Good philosophies are on display at Gnowsis relating to what they call the Semantic Desktop. The term seems ok in light of the Semantic Web which has taken its hold as a buzzword.

The MIT people talk about screenscrapers which are the web equivalent of the paper scanning concept.

I'm not clear what user interface I would find most useful. Right now I am just happy just to find open source. I wish it were not all Java but that seems to be the RDF language of choice.

Spreadsheets have been a very successful user interface. My hunch is that something which lets you draw your own graph links (like a vector drawing editor) might be worth exploration. But you'll always want table views for various purposes.

Historian James Burke of "Connections" TV fame now employs RDF-like graphs as a knowledge platform. Connected graphs of history and modern times can both intrigue and educate. Burke offers a real player video stream resume of his project. From what I can tell, K-Web does not use RDF, but has exactly the same idea. K-Web's eponymous term refers to connected graphs, not the WWW. The video lists Brain.com in its end credits. Wikipedia should do the job in RDF for Burke. Imagine using Wikipedia through an interface like Burke's onion. RDF would not only navigate node to node, which hypertext does now less elegantly, but also enable database queries and filtered views. I commend the K-Web initiative and hope it goes open source.

The overall GUI question is a field of study in its own right. Select your favorite descriptor: data visualization, visual complexity, information aesthetics. If we're talking about general RDF editing, for example, then traditional widgets won't cut it. They work only for narrow data types. We're discussing general graphs. Internet bookmark programs suffer from tree interfaces. What about bookmarks crossing several categories? The poor man's solution is setting short personal keywords in a comments field. I would rather use outright graphs or better, filtered RDF lens views.

Open Directory catalogues the cognitive software cottage industry under headings like concept mapping, knowledge management, and mind mapping. Open Directory offers huge RDF dumps if you need RDF test data.

One commercial offering which impressed me and others is written in Visual Prolog. That fact is quite interesting and the web site shows visualization possibilities. And before I forget, two open-source Java graphics libs are ZVTM and Wilmascope. The W3C's IsaViz uses ZVTM.

Mystickies.com looks like an ideal application for RDF. I don't know whether they use it.

Imagine visiting web sites with "favorite person" filters showing public notes by those persons.

Or imagine visiting sites with "subject" filters showing notes with actual semantic content relating to your request.

Or imagine querying a la Google for all notes with specified semantic tags or text content.

Semantic Web discussions