Fine, make me a blog

Among the largest impact research themes at the time of world-wide recession, the key subject is how to cope with mega company formations and e-government (digital government) projects that depend on the successes of information system integration. The current information system integration approaches such as wrapper approaches basically create combinatorial interfacing and/or combinatorial data conversion making the integration practically impossible because of interfacing explosion and/or computational explosion. A linear approach to overcome the combinatorial explosion is presented and discussed.

Well, he is either a cynic (knows the cost of everything and the value of nothing) or an enterprising researcher (pun intended) with a new look at system integration.  He calls it the Cellular Data Model, and this paper is his original position paper.  His personal website’s home page says it all in the title: “Welcome to unlimited worlds of knowledge.”

via Michael Bayne, a pretty smart Ph.D. student at U Dub.

Created by Dr. Philip Stanley-Marbell.  Dr. Marbell received his Ph.D. from Carnegie Mellon University.  His thesis was Programming Unreliable Networks of Computation.

Sunflower Hardware Emulator and Sunflower Hardware Platform are described in the thesis.

Evented I/O for Server-side JS

At the heart of our implementation is a new distributed synchronization protocol that we prove correct. Unlike several previous translations of synchronization abstractions in concurrent languages, we remain faithful to the standard semantics for Concurrent ML’s concurrency primitives. For example, we retain the symmetry of choose, which can express selective communication. As a corollary, we establish that implementing selective communication on distributed machines is no harder than implementing first-order message passing on such machines.

In this HotOS workshop paper, the authors argue that one should forget
Synchronous and Asynchronous models when designing “real-life”
systems, as none of these models accurately reflects the reality.
Their pragmatic proposal is to consider an Asynchronous model equipped
with a Perfect Failure Detector.

Perfect Failure Detector? In real life? If you know your FLP and your
Toueg, you know that PFDs cannot be implemented without support from a
synchronous network. However, thanks to some machinery, the authors
implement a PFD in a fully asynchronous system. The idea is simple: if
the PFD “thinks” that a process is dead, it turns this belief into
reality by surgically “killing” it.

http://research.microsoft.com/apps/pubs/default.aspx?id=80618

The tone of the paper upsets me. For example, they keep making very
rough approximation of key Distributed System concepts and of the
state-of-the-art in Distributed System. This is a very bad marketing
exercise, where the reader is asked to be an idiot that should buy
these arguments.

Nonetheless, the solution looks “ok” to me: it’s an “architectural”
solution to a problem that cannot be solved in full generality (as
predicted by the theory). So, that’s fine. Hence, if I take my
“distributed system” hat, I would say: that’s ok, do whatever you want
as soon as it seemingly works. Now, if I take my brand-new “operating
system” hat, I would ask for more measurements, scalability results,
real-life implementations, comparisons with traditional
asynchronous-based systems, etc.

Regards,

—
Pierre-Evariste DAGAND
http://perso.eleves.bretagne.ens-cachan.fr/~dagand/