about Eclipse and other rich client GUIs for the Semantic Web
In the last year, I was involved in many discussions revolving about the question of "how to make a semantic desktop gui". Semantic Web guis in general have to be dynamic, adaptive, generic, because the ontologies are changing and data of two ontologies can be mixed. This causes friction in software development, when all our frameworks and developers are used to relational databases or compiled Java beans that hold the information.
To gather my own thoughts for our NEPOMUK project I have prepared some slides to get an overview, a document summing up my view, and a screencast showing one framework.
To gather my own thoughts for our NEPOMUK project I have prepared some slides to get an overview, a document summing up my view, and a screencast showing one framework.
- creating rich semantic web guis - slides about haystack, dbin, fresnel, and some ideas for Nepomuk
- PSEWArchitecture connecting database, pimoservice and an Eclipse based user interface (PositionPaper) - a paper making the suggestion begun in the slides more precise, PSEW referring to the p2p semantic eclipse workbench.
- a screencast of myself using gnogno, our gui framework
|
leobard - 14. Jun, 15:52
|
|
- add comment - 0 trackbacks
Desktop Applications
> ontologies can be mixed.
I'm wondering how that works out usability-wise. Generalization is something geeks and software developer rave about, but an UI is often more effective and comprehendible if it concentrates one thing and does that thing right. I don't believe in "You can manipulate all kinds of data with this UI and everything is generated" before I actually see it working.
For example, how do you want to generate an UI from an ontology? Simply adding controls and input fields for all properties won't suffice to make a usable interface.
"This causes friction in software development, when all our frameworks and developers are used to relational databases or compiled Java beans that hold the information."
Really? We're talking about the desktop, right? Neither Java beans (or Java in general) nor rational databases are the building blocks of today's desktop applications. It's rather C, C++ and the like. Rational databases might become part of it but haven't yet.
desktop and ontologies
* hand-crafted guis are more usable than auto-generated
* Java is a minority on the desktop
When coming to "professional" software I disagree, especially with desktop ERP, CRM, Etc:
* MS-Outlook is build from standard components, the forms are a scripted assembly of components, the underlying basis is hardcore c/c++
* if you run SAP R/3 or Lotus Notes, most of the GUI is somehow auto-generated (not that R/3 is known for its top usability)
* We develop Java because we can run it on all platforms and its simple to program (comapred to c), so this is a development cost question
* Relational databases are sometimes embedded inside applications for easier programming.