[Hplusroadmap] Re: [OWW-Discuss] Tapping into open source / open access and doingslightly more

[Bryan Bishop]

On Thursday 15 May 2008, Antoon Goderis <goderisa at cs.man.ac.uk> wrote:
> Hi Bryan,
>
> Sorry! This got out before I finished it. I actually meant to point
> you to the data playground work and the work on Wings/Pegasus done in
> the scientific workflow literature.
>
> http://dx.doi.org/10.1109/E-SCIENCE.2007.72
> http://ai.isi.edu/div3/pubs/papers/gil2007wings.pdf
>
> Both rely on semantic descriptions of data types so you may find this
> helpful. Perhaps :-)

Yes. It seems to be like MapReduce/Hadoop in that it can take a job and 
split it up over distributed computing elements. I know that GNU Hurd 
is related, but Stallman's been running into problems for the past few 
decades. The MIT amorphous computing team is also related. These job 
scheduling semantic data formats (page 6 in the Wings paper) look like 
something that needs to be integrated with cupsd in the SKDB model:
http://heybryan.org/mediawiki/index.php/SKDB
But not only that, but there's the issue of amorphous distributed 
compilation, in that let's say you have a mechanical arm, and I have a 
mechanical arm in my fab-lab, and they're different although ultimately 
able to make the same things. How is this represented and taken into 
account? It requires some sort of "job compiler", the job format in the 
Wings case being one step of the problem. Any ideas?

- Bryan
________________________________________
http://heybryan.org/