harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From tian galaxy <harmonymail...@gmail.com>
Subject [GOSP]Proposal For OSGi
Date Thu, 02 Apr 2009 03:33:11 GMT
Hi All :

My name is Tian Yu ,I am a post graduate students from Shanghai Jiaotong
University .
I have just submit proposal for the summer code on Harmony OSGi.I want to
send all of you a copy of my proposal cause I know my plan could not be the
best one and need further improvement .So suggestions and comments of any
kind is welcome and will be appreciated .
I paste the detailed description below .

*Detailed Description:*

OSGi has a lot of advantages due to its ability to manage modules'
dependences and the service it provided,like you could register different
versions of bundles of the type simutaneously and let the user to choose
according to version numbers ,not the classpath.

Apache Harmony JRE implementation has already been designed to be
modularized and least coupled amongest the module .

If we could integrate the two together , we will bring great  flexibility to
the development and update of Harmony.And more importantly this will be more
attractive to users for it is more easy to use Since you could pick which
module, which version , which package to use .More than one version of the
this same module could also be used in one application.

So this is a very meaningful and interesting project .

I already have some draft ideas of how to implement this but any suggestions
and comments are welcome and will be appreciated.

I think our project could be further divede into four subtasks

Subtask 1: As we  all know there are already several successful
implementations of OSGi  ,the most popular ones are Equinox ,Knopferfish
,Felix .I think since Equinox and Felix are more related to Harmony ,we
could investigate on them to see how they could help us on this if
permitted. (I have installed both of them on my PC and  run some small
programs I wrote on them .I found both of them are easy to use , more
impressived by OSGi itself.)

Subtask 2: Though the modularization structure of Harmony has made this
project a lot easier than it should be , there also are  many tough issues
we should handle.The dependecies issue if one the main issues should be
concerned with.As the ability to handle dependecy is one of main attributes
of OSGi , the development of dependency structure is very important .I think
harmony may already has many information and conclusions of the mutual
dependency between modules .So the work is to develop the dependency
structure suits OSGi and Harmony best . I think its necessary to make a more
comprehensive investigation on the inner structure of Harmony .

Subtask 3: With 3 month it is hard to make huge changes .So we may borrow
some ideas of the implementation of J2EE ,as  to add an extra interface to
the class loader and make some adaptations .First of all , a bundle to
manage the OSGi itself is needed and should be loaded first .This module
should be able to manage the information of each bundle such as dependecy
,and should be able to provide services to core to OSGi (Fisrt to implement
the functions very core in the OSGi spec then add more). When we need extra
classes to load into the VM , we should load it through the OSGi manage
bundle:first examine which package , bundle the class belongs to  in the
list of registered bundles using dependency information then load the
specified package .

Subtask 4 : We could incorporate some module into Harmony to help the
development of OSGi programm ,as what the "bnd" do (if there is still enough
time ).This could futher reveal Harmony's advantage and feature as an
OSGi-ed JDK.

Time available :

At least 3 month(May-August) .Actually could start to prapare any time
before that  .More than 5 hours one day .

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message