For a few months now, I have been taking a look at software tools designed for kids' education. There are many of them available, but I have been focusing on the works of the community founded by S. Papert and Alan Kay, inventors of the Logo and Smalltalk programming languages respectively. Papert and Kay got inspiration from the pioneer figures of pedagogy Piaget and Montessori and they took constructivism as main leading path for their developments.
The main idea behind constructivism is that the kid (but also adults as a matter of fact) learn by doing and exploring concepts. Papert and Kay adapted this theory to the use of computers: computers are simulators of the physical reality and as such, they allow to learn by doing even for concepts for which a physical world experience would not be feasible. The Squeak eToys environment is a good example of this.
One thing that struck me is that these ideas lead to (among others) the following conclusion: give a kid a laptop loaded with open source software and an Internet and she can learn whatever she wants. This is not just theory. The best example of the application of this idea is the OLPC (One Laptop Per Child) project and its success.
The OLPC software is based upon a software suite called Sugar which is made of lots of so-called activities for different disciplines: math, languages, physics, etc. This open source environment allows a kid to learn things, not just by using the activities, but also by allowing her to look at the source code (how things are done).
The strength of this kind of environment is that it can be used at several levels which allow the user to use the optimal trade-off between power and easy of use.
I would like to see an analogous environment for Remote Sensing Image Processing and Analysis. The Orfeo Toolbox is of course a very good candidate for this, and many steps have been taken in this direction.
The work of architecture and integration of the best existing open source libraries (ITK, OSSIM, 6S, libSVM, etc.) has provided the kitchen sink, the power. The creation of bindings (python, java), has allowed to expand the ways in which this power can be accessed. Finally, the GUI-based application Monteverdi has made the point and click use of (a limited part of) this power possible. The availability of QGIS plugins demonstrates that other paths can be imagined to make OTB usable by non programmers.
However, there is still lots of work to be done in order to have a completely integrated environment where, from the same starting point, a user can navigate through the different levels of complexity and power. Even if this is not an easy goal achieve, I think that all the building blocks are available to progress in this direction. This may only be a (very) long term goal, but I think that the existence of this vision can help to federate a community and motivate people to contribute.