[OLPC library] Tentative Plans for Nepal's School Server and related infrastructure

[Bryan Berry]

Hey guys,

The purpose of this e-mail is to let the wider OLPC community know what
we are planning for the school server in Nepal's spring test school and
to solicit your ideas on what we can do better. Sulochan Acharya and I
are leading the work on the school server for the test school. If you
are interested to learn who the heck we are, skip to the end of this
e-mail.

Sorry for the overly long e-mail but I think Sulo and I and have not
been sharing enough of our ideas w/ the rest of the OLPC community. So
it was time for a brain dump
	

Our Ideas for the School Server
-------------------------------

1. Backing up Student data and sharing information:

We are looking to use WebDav to back up the individual student's home
folder to the school server.

Sulo will work this week on stripping down Moodle to bare bones for
sharing materials and general content management. Right now we are
considering using one acct for everyone. We don't have any existing
Moodle courses in Nepali so we aren't looking to deliver courseware
through Moodle.
 
Apparently, the guys at OLPC India are using drupal for content mgt. I
like Drupal but it seems to offer way more than what we are looking to
do: Share files, share projects, share documents.

How do we store team projects for later use? Perhaps the Journal will
allow us to do this in the future but I have the impression that it is
still under heavy development. Perhaps, the journal already handles this
well and I am simply misinformed. Please disabuse me of any false
notions.

Issues:
     1. Does the School Identity Manager support LDAP? LDAP support is
        essential for integrating existing content management systems w/
        the School Server.
     2. How do students share team projects for later use? For example,
        Laxmi and Arun are working on an essay together. Laxmi started
        the document and then shared it w/ Arun. Arun goes home and
        wants to work on the paper by himself for a while. He is
        connected to the mesh but Laxmi's XO doesn't appear in the mesh
        view. Can he still edit the document?


2. Digital Library 
There isn't much Nepali digital content on the Internet so part of OLE
Nepal's work will be to digitize poems, literature, histories, etc. and
store them "somewhere." Finding the right "somewhere," turned out to be
more difficult than we thought. We needed a repository that could
support multiple front-ends, w/ simpler front-ends for novice users and
more complex front-ends for the people that will load content into the
repository.

This may sound like overkill to a lot of people but that is because they
are accustomed to the abundance of materials in their own language on
the Internet and/or access to physical libraries. There are very, very
few public libraries in Nepal and Nepali-language materials on the
Internet are quite limited. We don't want to have to change out the
back-end repository 6 or 12 months from now if we choose something that
can't scale. Scalability is one of the reasons we chose not to use
Greenstone digital library http://www.greenstone.org/

We looked at Dspace and Eprints before settling on the fedora repository
server www.fedora-commons.org, not to be confused w/ fedora Linux. We
like fedora because it is a very powerful back-end repository that is
very scalable, up to 10 million objects. It is a true web service that
is decoupled from the front-end. We wouldn't be telling the full truth
if we didn't admit that we chose fedora in part due to the awesome
implementation of it by the Encyclopedia of Chicago
http://www.fedora-commons.org/about/outreach.php#video

We need to support multiple representations of a single object w/ in the
repository. For example, we need to allow users to access Nepal's
constitution as both a .pdf file and as an .xol bundle. We have to
support non .xol bundles because we want people w/out XO's to still be
able to access the library. Fedora lets us do this.

The problem w/ the fedora is that the off-the-shelf user interfaces are
fairly difficult to set up and maintain. Right now we are using
"Fez" http://dev-repo.library.uq.edu.au/wiki/index.php/Main_Page
It is a great front-end but was fairly time-consuming to install and
customize the interface.

Since the installation and maintenance of a fedora repository is fairly
complex we will try to use a centralized server and smart caching on the
school servers.

Matt Zumwalt of MediaShelf http://www.yourmediashelf.com/ has been
extremely helpful to us in working w/ fedora. I need to enlist his help
to figure out how to represent .xo and .xol bundles in the fedora
repository.

Issues for the Digital Library:
     1. We will need the School Server to aggressively support caching
        of requests from the library
     2. The .info file format for content and activity bundles is not
        XML. Ivan told me recently, "XML is a solution to a problem that
        doesn't exist." Perhaps he and Eric Raymond are right, but we
        need to a way to represent .xol metadata in the fedora
        repository  
     3. .XO bundles don't currently support Dublin Core metadata, the
        standard for digital archives. I will be discussing this later
        this week w/ SJ. For more about Dublin Core, read here
        http://dublincore.org/documents/usageguide/#whatis
     4. We need LDAP support on the School Server so users can access
        their stored searches and virtual "book bag" of previously
        accessed materials.


3. What our team needs

---------------
Hardware:
10 more XO's
2 active antennas

We really need about 10 more XO's and 2 active antennas. I will be in
Cambridge this Thursday and Michailis has generously already offered to
give me two. We really need more XO's to test out the mesh and for our
development team of 4 people. I believe that Sulo is filling out the XO
request as I write this e-mail.

Our development team does way too little testing of our activities on
actual XO's because we only have 2 working XO's at the moment. The
result is that we don't diagnose performance problems for our activities
until far too late in the development cycle.


About us:
----------
Open Learning Exchange Nepal is a local Nepali NGO that is part of the
grassroots OLPC community in Nepal. We have an agreement w/ the govt of
Nepal to help them implement OLPC at a Nepali school this April and
later in the fall at a larger # of schools. Sulochan Acharya and I are
working on the school server. Dev Mohanty of Nepal Wireless may also
spend a lot of time working on it w/ us.

There is also a development team at OLE Nepal of 4 people working
full-time on constructivist EToys for Math and English. They are working
hard to involve students, kids, and teachers in developing more
activities.

There is also the larger FOSS community that is contributing to OLPC in
Nepal. If you want to know more go to www.olenepal.org

There's more that we are working on but don't have to time to write
about right now. Thanks for reading to the end of this e-mail!

Bryan Berry
OLE Nepal
www.olenepal.org