Friday, May 25, 2012

Free flowing wiki domain models and racer kids

The spirit of the naked objects applications and frameworks is to have a platform that allows one to define a domain model with some structure and behavior. The platform then will just "act" out the domain model and behave as a specialized application.

Pushing that to the extreme, there's nothing more flexible and free-flowing as a wiki. How about describing the domain in a set of wiki pages and then "act" that out?

Well - I just put the idea to the test in this new website I am playing with: http://www.racerkidz.com

This is an example page or topic, describing a 'Club': http://www.racerkidz.com/wiki/Category:Club The "Categories" stand for concepts normally represented as classes in say a UML diagram. All the domain stuff you normally define in UML can be defined as you can see here, via annotations such as "[[Roles:User:Member,Coach,Fan]]" which imply that a User can associate to a Club in one of the respective roles.

This can drive out a lof of things, including:
- capturing information
- validation of info and associations
- display and behavior of the application etc


Wiki Surfers and Wiki Snakkers

The normal function of a wiki is to describe topics and allow one to surf them as topics are related. Simply because a topic is mentioned in another topic, we know it is related. If the 'domain model' told me how the 'kinds' relate, we can make that surfing so much more interesting and functional.

Snakking, which I explored in a separate project stands for quickly grabbing some information or data from some source, based on an XPATH-like navigation model.

You can see how easily we can extend the concept of XPATH from an XML file with tags that represent concepts to a wiki structure that represents... well, concepts. As long as we know the 'kinds' of the related topics, we can surf say from a Club to its Events with something like "Club/Events" with some sort of an... Wiki Path? What do we call this? WPATH?

Actually, you can see that at work right here: http://www.racerkidz.com/wiki/Season:OO_XC_2012/xp/Race/@date which will show you the dates of all the Races of the respective Season. Note that as of right now, most of those links do not even have their topics created... yeah, there's some magic going on!

DSL vs Wiki

You could say that the tide is now firmly in the DSL's court (Domain Specific Languages). How about wikis instead?

The parallels are I think rather obvious: DSL is mostly structured with maybe free text comments as annotations while this approach in a wiki is mostly unstructured free text with structured annotations here and there...

Putting it all together

In this -I think- ultimate of incarnations of naked object models, the domain model grows as needed.

The domain model mingles domain artifacts (properties, types, kinds, categories, relationships) with free-flowing descriptions, eskeqing the need for conventional documentation. I mean everyone is able to read a wiki topic, even if they have to skip those funny annotations, right?

Filling out the actual data is also free-flowing from this domain model. At the very minimum, no extra syntax is needed, since one can simply link out and connect the topics, then the surfers and snakkers do the mining and custom logic.

Extracting data in different easily customizable ways is done easily via snakkers. The bit about behavior is a little harder and we'll explore that some other time.


Now what?

Where can we go from here? Well, there's many obvious directions.

The Semantic Wiki initiative implies the need for interaction between the separate wiki 'islands' - that implies an API.

Collaborative directories and/or indexes like DMOZ are another way to organize information. There's clearly a connection there that deserves exploring.
In fact probably a majority of new websites are the same: trying to index and organize someone else's data, drive some traffic and drink away the ad moneys. How do we instead derive new information from existing one? Or make it functional and useful at least?

Throwing some RDF, OWL and others into the mix.

Exploring the behavioral aspect - some scripting of some kind, reactive or not?

Including comments in the surfing and snakking.

Marrying this obvious graph with some graphical, visually striking presentation? My favorite is TheBrain?


The RacerKidz website

It's been fun trying out newer technologies like Play, Mongo and Scala to explore these concepts and I think the site may be useful on its own. I certainly hope for more and more active users, not only to create the oh-so-lacking contextual content about local competitive sports but to also continue exploring the possibilities of this approach.


I am generally depressed at the stupidity and laziness embodied in many a website and software today, so expect this website to become as smart and easy-to-use as I can make it.

So, if you or your kidz are interested in any competitive sports, please join in and play with it.

Cheers,
Razie

0 comments: