Hi, This is why I had asked the question of VM binary layout compatibility on the "Single Module" thread. ( no link ) :-) What Oliver is saying sounds reasonable to me. Please see below..
On 5/18/06, Oliver Deakin <[EMAIL PROTECTED]> wrote: >>Andrey Chernyshev wrote: >> Well, may be I'm missing some simple points here, but... >> Is there any specific reason not to bundle a "Harmony VM" (let's >> forget about DRLVM for now) into the HDK? Is this just a matter of >> choice between SableVM, drlvm, e.t.c.? >> I just thought it could be convenient for developers (as well as for >> Harmony users) to take a complete workable JRE without an additional >> need to combine something together. >I agree it would be useful to have a complete jre available, but I >wasn't sure >how we would pick the VM to use when we potentially have 3/4 VMs >(if DRLVM is accepted, and once the classlib adapter is completed) >capable of >running with Harmony classlib.
It seems likely that at some point, we will have to converge on the concept of full Harmony releases/distros( CLasslib + VM ) that we will be able to test, stress test, perf analyse and support( field bugs on ) regularly. This will probably require us to pick and bundle a VM of choice with our class libraries.
Perhaps we could just produce a jre for each VM once they are up and >working? >I just thought that if we had separate classlib and a VM jre's which >both contained >a predefined and matching directory structure, then the step of >unpacking them i>nto the same directory to get them working is fairly straightforward. This uniformity in binary layout required of the VM's sounds sensible.
This way, whatever default VM we roll out the distros with, any user/developer would be able to overlay the classlib jdk/jre bits with the VM bits of choice, which may not be the defaultone packaged. We may also organize the distros ( full Harmony, Harmony classlibs, VM1, VM2 etc. ) to enable this. Thanks, Rana
