Lots of universities, schools, governments want to use software in education
Kids around the world, particularly in poor countries, need interactive, free materials
The current method of education in much of the developing world is:
Teacher Speaks, Kids Repeat
An Education Myth ” Rote Learning is good for you, look at all those successful asians in the US”
Large percentage of those asians attended private schools or special public schools that look a lot like western schools, but often more rigorous
Those who attended public schools have largely succeeded on their own merits not because they had to memorize whole books as children
The Missing Teachers Who is going to teach all these kids algebra?
The Missing Teachers Talented teachers won't move to rural India, Pakistan, Nepal, Africa for another generation. Vast majority of people who know maths, English, good writing emigrate or go into private sector NOT teaching in a village
We need to fill this gigantic learning void
We need to empower good teachers and accommodate those learners that don't have access to good teachers
And You Can Help!
Bryan Berry, CTO of OLE Nepal, had interesting experience on one visit to a Nepali school. He asked a 3 rd grader what the sum of 3 + 1 is. The child enthusiastically responded “4!” Bryan then asked the child what the sum of 1 + 3 is. The child had no idea. The child had memorized addition problems but had no actual understanding of arithmetic Anecdote
Problems with Current Approaches
Most interactive educational software today suffers from one or more problems:
Tied to a particular platform
Pygame +PyGTK (gnome)
Very small educational dev communities – gcompris, KDEedu, Squeak
But Flash is the Standard! ” Flash is the standard for developing proprietary educational software” But the open-source Flash community is small and not very active Even Worse:
You can't show off your creations through the browser
Separate markup, presentation, and code? Nope
Very, very few programmers in developing countries know these technologies
Lots of programmers in developing countries know these technologies due to outsourcing boom
Great unicode and font-rendering support
Low barrier to entry, new coder only needs text editor and browser to get started
A Lot of Work Ahead
JS still lacks standard high-level drawing functions
Browser performance is very inconsistent, particularly for certain kinds operations
Lack of good i18n mechanisms that don't require a server
Road to Karma 1
Open Learning Exchange Nepal (OLE Nepal) is a Nepali NGO implementing the OLPC project in Nepal
2000 XO laptops deployed so far
More importantly, several hundred teachers trained how to use XO's in classroom
Educational content aligned with curriculum critical to project success
Road to Karma 2
OLE Nepal has been producing open-source educational software for 2 years now
Tried squeak and pygame + pyGTK but development too time-consuming
OLE Nepal has been using Flash for last 18 months and happy with it, BUT . . .
Lack of External Support
To date, not one single developer outside of OLE Nepal team has submitted a patch despite our best efforts
Relative to the huge # of devs working w/ flash
IRC, mailing list traffic low
Limited open-source tooling and supporting documentation
The Open-Source Flash Community will likely remain small
Lack of External Support Despite the great work of the free software community on projects like Gnash and the GameSWF library, there is very little in terms of sharing and collaboration of free software Flash projects.
Insult to Injury OLE Nepal has to use images for Nepali text because Flash Player 10 does not render unicode indic fonts properly
All the Same OLE Nepal has created some amazing content using Flash It will take a ton of work to implement the same functionality in html5+js
OLE Nepal's Mission To provide comprehensive, high-quality education to every Nepali child. Must cover grades 1-8 with teacher training and interactive, high-quality, digital content aligned with national curriculum
That's a lot of Content OLE Nepal will never, ever, ever be able to create all this content on its own Creating an international movement around standard tools is a more realistic goal
There are lots of people around the world interested in creating this content
They just need a common, open, ”easy” toolset
OLE Nepal needs to move from Flash to openweb technologies => Karma
Karma has generated more interest in just 4 months of existence than did 2 years of flash work
And there is only one basic example . . .
A SugarLabs Subproject
Karma is a subproject of Sugarlabs
Started as Googe Summer of Code project – thanks Google!
But Karma lessons designed to run within any html5 compliant browser, not just within Sugar
Code is MIT-licensed, content CC-BY-SA
A SugarLabs Subproject
OLE Nepal and others in Sugar community realize that to maximize amount of content created, karma can't be specific to a particular desktop environment
Who's Working on It?
Created by Bryan Berry, CTO of OLE Nepal
Main contributors so far
GsoC student Felipe López Toledo
Christoph Derndorfer, co-founder of OLPC Austria
A group of passionate people who know relatively little about web tech
Who's Working on It?
OLE Nepal has a full-time paid development team consisting of 3 teachers, 3 graphic designers, and 6 programmers
Why aren't they all working on Karma?
The government of Nepal has contracted OLE Nepal to produce a certain amount of content according to a set schedule. Shifting all development resources to Karma would jeopardize that schedule
Everything must work both offline and online
Only dependency should be a browser with good html5 support
Browse and Surf under the XO
EASY i18n, gettext doesn't play well w/ MVC
Text in markup needs to be easily localized
We won't have server-side templates, forget Gears
Helper scripts like the django admin.py and rails ”rails” script
Requirements for Toolset
Use only open-source components
Libraries for high-level drawing functions, audio manipulation, common bits of lesson functionality, etc.
Low Barrier to Entry
Use webdev paradigms
Convention-over-configuration a la ruby-on-rails, django
Requirements for Toolset
Web dev paradigms
Separate presentation, markup, and code
What We Don't Need Exciting 3D Graphics High-quality video Simple 2D is enough And this needs to run on cheap computing devices And 3D doesn't make that much of a difference, think gameplay, gameplay, gameplay
How You Can Help
High-Level libraries to ease manipulation of html5 <audio>, <canvas>, <video>
Better i18n mechanism
A JS rockstar to step up and advise the Karma project
Optimizing animation performance in the browser
Help with web workers
Help w/ local storage to store student progress
without server-side template
Capture text in the markup and html attributes like <img alt=””> and <meta tags=””>
Logical Layout of a Karma Bundle
The base unit of karma is the ” lesson ”
Lessons are organized by grade, subject, and optional time component in a Karma bundle
Each lesson includes
Lesson Plan and Teacher's Note for teacher
Tutorial – introduce concept
Exercise – do exercises to test knowledge
Optional game – more free form opportunity to work with ideas in the lesson
Anatomy of Karma codebase
Helper scripts – presently using narwhal + rhino
Starting page – Chakra
Anatomy of Karma cont'd
A formal specification (under rapid development): wiki.sugarlabs.org/go/Karma/Specifications
Set of conventions for most common configurations very much in the style of ruby-on-rails and django
No server-side components or even server-side emulation like Gears
The Kurrent Karma Plugin
Presently not using SVG. When should we use svg? When use canvas?
The content must run under OLPC XO-1
SVG tests crashed the XO-1
Canvas tests works better under XO-1
Created own plugin: jquery.karma.js
Karma Animation Felipe tried RaphaelJS (http://raphaeljs.com/) demos to test SVG animations on the XO but it either crashed or suffered horrible performance Since canvas is low level, better performance … But it's really easy to manipulate SVG Objects Karma doesn't support SVG as of version 0.1 but we would like to in the future
Karma Animation 2
Need something simple and high-level
processing.js great but complex to use
It uses processing syntax.
It uses a parser, so it’s slower than JS.
It’s not designed to work on low-power machines like the XO (433 Mhz processor, 256 MB RAM).
Karma lessons must run on the XO-1.
Default browser: Sugar Browse ( based on Gecko )
Experimental: Surf ( based on webkit )
No problem rendering HTML common elements
Canvas rendering may be slow Quadrilaterals demo
Several lag on Browse (unusable)
Works nice on Surf
Browser Optimization 2
It’s preferable to spend cpu than spend memory.
Emphasize rendering speed over quality
image-rendering / firefox 3.6 alpha
webkit has nothing
An animation core as web worker?
Trying to combine actions with predictable behaviour.
it's not possible to send functions as worker messages, so, it's hard to use callbacks.
how to share memory (variables) between the root document and the worker thread without copying that data?
We need to store results from exercises in Karma lessons
Later synchronize them with moodle or another learning management system next time student can connect
i18n is too hard
Current i18n mechanisms require a server-side template – smarty, erb, django's? –
Or putting all your strings in js code
Gettext's _(””) works for js code but not for markup
We need something more intuitive
Specific i18n Needs
Need to translate text in markup not just stuff marked _(””), even including inline markup
Need to use native numerals – Arabic, Hindi, Nepali use different symbols for digits
Need to change out audio, images, and even numerals per locale
Gettext should translate all strings in <body> unless marked explicitly otherwise – perhaps using lang attribute?
The markup still needs to be valid html5!
Plus, need to translate certain attributes such as <img alt=””> and tags in header such as title, meta
CSS background images and fonts
Doing Everything on document.ready()
Since we can't use something like a server-side template, all localization has to happen in function called on document.ready()
Switching out images
Will this actually work for a complex pages on machines with limited resources like the OLPC XO-1?
Inline markup is a headache
<div>the <big>Highest<big> score is <label id=”highScore”></label></div>
We need grab everything between <div> . . . </div>, put it in the .pot file, and later write back the translated HTML on page load, not just text
Anyone got a better idea?
Could you help?
Image sprite support
Mechanish to resize the whole html content when resizing the window
Could you help?
Html2po (supporting inline elements)
Could you help?
using only JS
Only under XO?
Have to rely on frameworks like Telepathy?
Early ideas on http://wiki.sugarlabs.org/go/WebCollab