[lsb-discuss] LSB conf call notes for 2008-05-28

[Joseph Kowalski]

Thanks for the clarifications.

Jeff Licquia wrote:
>>> Ron: different JRE compatibility?  George: IBM JRE is certified.  
>>> Kay: high level of compatibility.  Darren: his experience also.
>>
>> Isn't that the point of the LSB?
>
> Sorry.  "Certified" -> "Java trademark compliant", meaning Sun is OK 
> with IBM calling its JRE "Java".

I'm pretty sure the Sun (who owns the trademark) allows certified 
implementations to call themselves Java.  I believe there are rather 
explicit statements prohibiting the use of the name Java for 
non-certified implementations  (certified in this case means JCK).

>>> Jeff: DBus?  George: not sure; depends on the implementation. There 
>>> is a DBus implementation for Java.
>>
>> I'm really surprised to see "DBus" and "Java" in the same paragraph.
>> Is this a desire for extra classes for access the DBus, or is this 
>> something
>> else?
>
> In general, D-Bus is becoming the standard IPC mechanism on Linux for 
> lots of things.  GNOME and KDE have both announced (and in KDE's case, 
> implemented) that they are moving to D-Bus for this, so I expect it to 
> become interesting on Solaris, too.
>
> More importantly from our point of view, the a11y people are busy 
> porting AT-SPI from CORBA to D-Bus, so in the future the 
> aforementioned a11y bridge will need to know how to do D-Bus.


I can still read this two ways,... maybe both:

    It is desired that Java implementations use D-Bus.

    Java applications need/desire D-Bus access methods.

????

>> Lots of things need platform-specific code - called via the JNI.
>> That's not the same thing as "compiled for each JRE" (ally).
>
> So JNI should be stable enough from a binary ABI perspective that a 
> JNI module compiled with, say, Sun JDK should run on IBM's JRE?  
> That's good to know.


A JNI module has a baroque, but very well defined interface with the the 
Java Application.  A lsb-compliant JNI module can use the well defined 
interface (known as the LSB) to play well with the platform.  As a 
matter of fact, there *needs* to be some verbiage (in the LSB) that JNI 
modules associated with the LSB conformant Java application are LSB 
conformant themselves.  (There *could* be a significant effort need to 
generate tests to validate those words.)
         

>
> Might it make sense to define that binary ABI as part of the LSB?


Yes it might...

    1)   I think this might just come out of the "Java Specification".

    2)   As a process, I think we should get the interfaces for "Pure" 
Java applications defined for the LSB first, and then worry about 
Java/JNI applications.  Just a "divide and conquer" thing.

- jek3