[Mandala-devel] Welcome !

[Pierre V. <ei...@ne...>]

Hi everyone !

I've commited a lot of things this week-end.

Here the list of good news for you who want to know more about Mandala :

I've modified the Makefile so the command

make mandala_docs

will generated all class files and the javadocs.

All files generated by javadoc are in the MANDALA_DIR/docs directory where
MANDALA_DIR is the directory where you have made the cvs checkout (on my UNIX
box, it is ~/java/Mandala).

Hence, after the execution of the previous command, you may give to your
browser the following URL :

file://MANDALA_DIR/docs/index.html

You can click on the link 'Overview' which will give you perhaps a better
understand of what Mandala is and what it provides.

Then, i've also made a TODO list (MANDALA_DIR/TODO) which is up to date now. If
you had watched the content of this file before, please, excuse-me it was in
fact the file of another project (JTOE).

For convenience, i copy the content of this file here so we can better decide
who will do what.


**** TODO content ****

+ GENERAL

	- Making : 

		* try to guess the MANDALA_DIR variable in Makefile.common
		* Generate better rules for RMI stub (always regenerated even
		  if already done !)
		* use a ./configure to guess the compiler (javac, jikes, gcj)
		* Defines Makefile to run also on DOS based platform
		* perhaps use another tool like Ant !
		* Add 'jar' rules in Makefile to generate 'mandala.jar'
		* Add 'tgz' rules in Makefile to generate an archive containing
		  only the jar file and the javadoc for end users.
		* Add 'install' rule in Makefile.
		* Provide 'rami', 'jacob', 'util' rules which just build the
  		  necessary subpackage.
		* Provides a 'rami-tgz' rule which build RAMI (jar and doc) for
		  end-users who just need this package.

	- Running :
		* Generate scripts for DOS based platform (in bin, and in
		  bin/rmi)

	- Docs :
		* Correct Javadoc comments (bad english, grammatical errors,
		  ...)
		* Design a good web site ;-)
		* Create a good Mandala Logo (I've already a JACOb logo, and a
		  RAMI logo. They probably sucks, so if you have better skills
		  than me in graphics design -- which is really not hard -- be
		  free to propose your logos !).
		* Create a Mandala-specific Doclet to customize the generated
		  javadoc.
 

+ CODE

	- Rewriting - updating
		* Rewrite packages 'mandala.jacob.remote.test' and
		 'mandala.jacob.remote.protocols.rmi.test' so they use the
		 JUnit framework.
		* Update the jayac compiler so it can run asynchronously (using
		  RAMI of course), and can be distributed (using JACOb of
		  course). The new jayac must provide classes for any super
		  classes, interfaces and superinteferaces that the given class
		  inherits/implements.
		* Write a new JayacFromSources which runs from sources and
		  not only from classes.
		* Define a good semantic for Callback in
		  the AsynchronousReferencePair class.
		
	- New features :
		* Add an UDP implementation of the JACOb framework.
		* Add a TCP implementation of the JACOb framework.
		* Add an UDP implementation based on java.nio (>= JDK 1.4)
		* Add a TCP implementation based on java.io (>= JDK 1.4)
		* Add a Myrinet implementation of the JACOb framework.
		* Add an HTTP implementation of the JACOb framework.
		* Add a CORBA implementation of the JACOb framework.
		* Add an SCI implementation of the JACOb framework.
		* Add any other implementation you think good of the JACOb
		  framework ;-)
		* Add a GUI to JACOb. Objects may be inserted, removed
		  dynamically from this GUI. When an object is clicked, a list
		  of its method must appear, and any one of it may be invokable
		  by a way to define (parameters ? Perhaps using a beanshell).
		* Add services to JACOb : security is the major lack in JACOb,
		  hence it is the priority. A design for services must be
		  discussed. Should services be stocked objects ? Should they
		  be special objects ? Other services are monitoring,
		  persistance, transaction, ...
		
Probably many other things to do...

**** end of TODO content ****


I would like to work on the asynchronization/parallelization/distribution of
the Jayac compiler from classes, not from sources which is another -- probably
big -- task. If someone is interested in this task too, i can :

        - release this task and take another free one,
        - propose him to co-develop with me,
        - join another developer (or maybe a team of developers) which are
        already working on a given task.

I hope you're all interested since as you see, a lot of work as to be done.

Ha, i forgot to say that i've already some projects that uses Mandala (but a
really old version : the prehistoric first version of JACOb !) in a distributed
computing context.

They are : 

        * Travelling Salesman Project,

        * Distributed computing of the number Pi to the n th decimals with a
          central load balancing and fault tolerance algorithm.

        * Spectral analysis of waves sound with a general purpose central load
          balancer

I've a contact which would like to use Mandala for the purpose of the
parallelization of a research application in the context of biological cellular
behaviour simulation.

I also would like to propose at least two frameworks based on Mandala :

        * a mobile agents system (like Voyager, Aglets, Concordia, ProActive)

        * a general distributed computing framework. The idea is that end-users
          may lease their CPU/IO/Memory, for a period of time just by clicking
          on a web link (an Applet). But, the computer leased, is not dedicated
          to a sepcific domain, it can be used for one period to compute alien
          presence (seti@home) and a few minutes later, to do some meteo
          related computing.

I know such frameworks exist already (distributed.net for the last one for
example) but i believe that since Mandala is based on concepts (asynchronnous
references and active containers), it may provide more naturally good
fundations for such frameworks to be build on, with mathematical, proof of
correctness caracteristics (active containers are modelized in a formal model
called pi-calculus).


Suggestions, ideas, remarks, critisms is welcomy ;-)

Regards. eipi.

-- 
No one equals eipi : -1 = e^(i.Pi)