Raw

The last couple of suggestions were a bit dull really, so to compensate here's something more fun. Basic idea is : take an RDF toolkit, add a GUI, wrap it up into an easy-to-install bundle. Everything you need to start learning about the Semantic Web, right up to developing ontologies and applications.

Some precedents : Lisp in a Box (Common Lisp + SLIME IDE/emacs); Userland Frontier (content management system); Squeak (Smalltalk environment).

Lisp in a Box is essentially an IDE, only really of note for the way it's other bits and pieces co-packaged. Squeak is totally immersive, and in my opinion a wonderful playground for UI work (assuming you don't mind a bubblegum colour scheme), but has quite a steep initial learning curve (if you're over the age of 9).

Frontier is closest in its level of self-containedness to how I picture SWAB. Ok, so that example is a fairly closed-box CMS, but hopefully you see the idea. Frontier is also ancient, but Les Orchard still found things to like - note in particular his phrase "instant gratification". Exactly.

The other week, talking to Chris Langreiter and Tom Fuerstner, they both mentioned how they'd been influenced by Frontier. Tom in fact said (probably paraphrasing badly) that after an early start with computers, he'd dropped out for a decade or so. It was Frontier and the Web that got him interested again. Frontier got an awful lot of things right back then, but the overall application architecture now seems dated : content-oriented, standalone, localised, a platform in itself, Web 1.0. The platform of Semantic Web in a Box will be the (Semantic) Web (2.0).

There are quite a few tools appearing under the umbrella of Semantic Desktops, but rather than SWAB being a tool for immediate practical use, it would be a toy which encouraged play (and perhaps development of practical apps). It would be best if you could do stuff right out of the box, with zero prior knowledge.

So here are a few possible features :

• trivial install (semweb-box.msi/semweb.deb/semweb.jar)
• text editor (syntax highlighting for Turtle/SPARQL would be nice)
• RDF store - file-based would be fine, though a no-install RDB setup would be fine
• HTTP client (not necessarily with HTML browser, but scutter capability would be good)
• HTTP server
• SPARQL
• inferencing - RDFS, OWL DL, rules, the more the merrier - as long as it didn't make the UI look complicated
• scripting language - I played with Jython + Jena a couple of years ago ( screenshot), good fun
• XSLT/GRDDL
• Modular, pluggable architecture - plugins only needing one-click install (like Firefox plugins)
• Sample apps - at least one content-oriented (e.g. blog/aggregator/wiki/Atom Store?), one data oriented (e.g. address book, PIM?), one comm-oriented (IRC bot? - handy for inter-box chat!)
• docs

Having a HTTP server also suggests that a 'headless' server-side box install could be worth exploring, and plugins prehaps running standalone.

There's a lot on this list, but a good 90% is covered by virtual every RDF toolkit out there (plus an extra lib or two), it's mostly a matter of GUI-writing and packaging. I'd strongly caution against trying to include too much at the start - the project in the screenshot above stalled because I was trying to include too much ( other screenshots). Bare-bones + pluggable is the way to go, I'm sure.

I mentioned SWAB on one of the IRC channels the other day - response was basically : cool, but X can already do that. Yeah but no but if Auntie Maud asks about how to get into the Semantic Web, see what happens if you tell her "download cwm (/Jena/Redland…) and have a play". GUI + packaging please.

Leigh Dodds' Twinkle SPARQL query tool is pretty close to the kind of UI I'd like to see : uncomplicated!

Other SWEO Suggestions.

[Danny]