Byzantine Askemos Language Layer

ABHistory :: pdf :: context view

inbound links:
History & People

History & People

The work on this project has covert a good part of my life. But for almost a decade it was just an idea how a global computer should work.

Ben Howell Davis
The Global Computer is not an object it is a property.
During the year 2000 I wrote a prototype to illustrate how a single peer could work. While released to the public, there was no reason to exaggerate about the state of development.

Hans Jürgen Stüber and Eberhard Richter nevertheless understood the value of the idea. In a fruitful cooperation their companies deployed this evolving prototype for the next ten years to build research prototypes of several applications under the assumption that fully working Askemos would enable them to work for real. Several students and employees participated. (TODO:dig up all names. Raphael Heinrich, Christian Schaller, Tom-Steve Watzke...)

A the same time me and for some time Peter Hochgemuth refined the prototype to actually become a P2P network. Furthermore a lot of code went into features just deemed practical. (E.g. a WebDAV file system to safe applications source code and transparent sqlite replication.)

Note by the way: This situation lead to an rare kind of testing. Every change to the software was tested not only by unit tests but put to work under the hood to unaware application programmers -- effectively in between two mouse clicks.

Some scientists became especially interested during these years: e.g., Hans Gert Gräbe, Klaus Bastian. They asserted so far, that the idea is solid (here a -german- lawyers expert opinion) while the implementation is just a prototype... (TODO find most of the reviewers.)

In 2010 we've been able to pass two milestones: a) "SQL working" b) "Independence from language implementation" (the code can be compiled by at least two compilers now). By the end of 2010 the development network - running many applications (including ordinary customers websites) - became stable enough for independent deployment.

Summarizing the state by end of 2010: applications which need a single transaction per request - that is "as normal web sites" do - are easy. However apps, which incur sequences of transactions on independent objects (notably when those objects are supported at quorums which have none or a partially intersection only) are still more art than engineering.


Last modification: Fri, 24 Dec 2010 17:24:01 +0100

Author(s): jfw,

Document number A26b5619be8d5e3348cca356acfc8efea delivered to public at Fri, 24 Nov 2017 13:37:16 +0100

short comments

Add