Network2: Difference between revisions

From OLPC
Jump to navigation Jump to search
(..)
(4)
Line 1: Line 1:
{{TOCright}}
{{TOCright}}
Last updated: [[User:Mstone|Michael Stone]] 23:43, 26 July 2009 (UTC) | [[Network2/Paper|paper version]]
Last updated: --[[User:Sj|Sj]]&nbsp;[[User talk:Sj|<font style="color:#f70; font-size:70%">talk</font>]] 04:38, 28 July 2009 (UTC) | [[Network2/Paper|paper version]]





Revision as of 04:38, 28 July 2009

Last updated: --Sj talk 04:38, 28 July 2009 (UTC) | paper version


Sugar's view of realtime collaboration (see network principles) relies on a robust network stack, which does not exist today. Its intended user experience relies on something stronger: a network which fails in predictable ways, and can be manually overridden in order to get on with a desired behavior of Activities.

This essay summarizes an attempt to work out a simple way to realize this sort of network experience, with existing software and hardware, and to demonstrate the sort of thinking which might help other parts of the system contribute to the same user experience.

Specific goals from a discussion with Michael:

  1. "I want to chop out 2-3 layers from the current six-layer network stack."
  2. "I want to name my machine heap.mstone.info and have that identifier work whether I am sitting under a tree with a mesh link to you or on a wide-spread network."
  3. "I want to be able to go to /etc/hosts and type in the address I want in case it can't be found otherwise, and have it just work"
  4. "I want a design that never mandates single points of failure."


Longer essay

Usability

There are many ways that children involved in the OLPC effort might fail to benefit from their involvement because of network design and implementation failures such as lack of interoperability, efficiency, and usability.

In order to address these risks, we have tried to explain a context of work, a number of designs, and several implementations that we think will help mitigate these threats in both wikified (below) and paper forms.

Unfortunately, providing truly dependable software is a challenging task at best. Fortunately, there are many ways that you can help out, both generically, particularly, or via role-based scaffolding according to your preferences. Finally, if you are interested in speaking with networking people, know that they are readily available.

Context

This network design effort is growing in the fertile ashes of previous network and collaboration attempts. Consequently, it proceeds from previously realized Network Principles according to a multi-faceted statement of purpose, a prioritized list of network scenarios, and a collection of architectural quality principles wrought from the aforementioned previous efforts.

Details

Networking is intimately related to all aspects of the Sugar experience, including both usage and creation. Here are some pages describing many aspects of this experience. (Caveat: When judging, please also note that the design is not yet complete in several important respects, as indicated below.)

Usefully complete

Naming and Internetworking
We intend to use DNS and IPv6 for naming and internetworking.
Diagnosis
Our quality principles mandate that we provide fault diagnosis procedures even before providing implementations!
Experiments
We have begun experimenting with naming and tunneling technologies like dnshash and openvpn.

Stubs

Activities
We provide advice for activity authors on special factors to consider when writing networked or collaborative activities.
Security
Thoughts on where we might want to go with communications security.
Dynamics
In order to efficiently study scaling costs, we maintain bandwidth and latency models for our most important links, protocols, and implementations.
Self-test
A logical extension of good manual diagnosis procedures is the creation of good automated diagnosis procedures.
Future work
As with any ambitious project, there's always more to do!

To Be Done

Management
How do we dynamically configure links and routes?
Integration
How do we get this merged?


Contributions

You can contribute to the education received by hundreds of thousands of children this year by:

writing software
Review the documentation cited above, then bring your questions and patches to the sugar-devel mailing list (subscribe).
refining the context and design
Did we miss an important threat (e.g. to availability)? If so, please work with us to fix our model.
Alternately, if you have expertise in a related field like mathematical modeling (how far can this scale?) or usability (how should we inform users of failure?), please improve our theories and recommended practices.
breaking assumptions
Networking is proven both in the mind and under fire. Here's your opportunity to crank up the heat.
organizing other people
Many people are capable of improving the networking ecosystem but for the lack of some critical resource like knowledge, motivation, or criticism. Find and provide the missing piece.
spreading the word
Many of our networking ideas are transferable to other operating systems and environments -- particularly to other Unix-like machines. Help port our ideas or software to another platform so that others can benefit from them and can help us improve them on their own terms.

Procedures

Some day soon, we'll try to write up some simple procedures to ease the task of making the networking contributions described above. Ping the sugar-devel list (subscribe) if you want this up.

Thanks

Many people, both named and anonymous have contributed to the network ecosystem containing the XO and hence to the quality and power of the education received by hundreds of thousands of kids this year. If you or your organization would like to be recognized for your contributions, please add your name and affiliation to the Network2 credits page along with a brief description of what you worked on.