Hello all,
 
 On Sun, May 15, 2005 at 06:42:16PM -0400, Patrice Le Vexier wrote:
 > Dear all,
 > 
 > correct me if i'm wrong, but, to make a J2SE platform, we need:
 > 
 > - a compiler (javac)
 > - a JVM (maybe a simple interpreter at first time, then a jit/wat/xxx of the
 > death)
 > - class library (classpath as good bet)
 > - some tools (appletviewer, keytool etc)
 > - an installer
 > - ...
 
 True, but from an architect's point of view, I believe we should look
 closer at some fundamentals WRT requirements and at a common
 implementation layer. Questions such as 
 
 a) Do we have a go from Apache and Sun Legal to actually pursue this  
    project? (Noone wants to end up in a legal battle somewhere down the
    road).
 
    ==> Geir? Your ball?
 
 b) Which OS's should be targeted?
    (This gives info for a baseline to installation and OS 
    abstraction layers.)
 
    Given experience from some large open source projects such as Debian
    GNU/Linux, targeting a multitude of OS'es can slow down the project
    considerably - while at the same time creating a de-facto standard in
    strength of its many platform options.
 
    ==> Is it sensible to start with the OSs [Windows 2K/XP, Linux <pick
        libc version>, Solaris, Mac OS X]?
 
 c) Could we augment the proposed architecture in
    (http://people.apache.org/~geirm/harmony.jpg) by introducing an OS
    abstraction layer (OAL) to failitate porting? (Similar to the
    strategy used for GCC porting.)

    Placing a quick sketch of this on 
    http://www.jguru.se/images/harmony_OAL.jpg
 
 d) What strategy do we pick WRT JDK tool implementation? I would
    recommend that we create a common library that can be used by all
    tools. That way, we can minimize the porting effort to different
    OS'es and have a single library to distribute instead of a splay 
    of smaller ones. 

    Placing a small sketch of this on 
    http://www.jguru.se/images/ctl.jpg
 
 > I know we're at the very first stage of the development and a lot of stuff
 > are currently actively discussed, but is there nothing of this list we can
 > begin to code, maybe just as proof of concept, or to try some solutions ?
 
 Since Harmony is potentially an *extremely* important project, I
 recommend that we stick to the Apache development process. We will have
 quite a lot to code soon enough.
 
 Patience. 
 
 ;)
 
 ---
 // Lennart Jörelid, Senior J2EE Architect
 // jGuru Europe AB
 // [EMAIL PROTECTED]

Reply via email to