Re: [certi-dev] CERTI Java Binding

[Eric Noulard]

2009/11/20 Gotthard, Petr <address@hidden>:
> Eric, Andrej,
>
> Just for clarification:
>
>> >  On a related note I am not sure whether it is possible to use UNIX
>> > socket natively in Java without JNI. I will look into it later but I
>> > am pessimistic because of what I have seen so far.
>>
>> We already did look into this,
>> It is not possible in portable manner
>> (unless it has evolved since 1, 2 years but I doubt it because
>>  as far as I know there is no interest in this feature).
>>
>> You have to use JNI.
>> That's the second reason (after the Windows port)
>> why we did optionnally use TCP socket between Federate and its RTIA.
>
> With respect what we discussed past months
> http://lists.nongnu.org/archive/html/certi-devel/2009-05/msg00086.html
> http://lists.nongnu.org/archive/html/certi-devel/2009-10/msg00008.html
>
> ...and with your current understanding, should the JNI be used to implement 
> the Java CERTI binding, or is the Python message generator still an option?

Python is definitely the BEST option.

The question form Andrej is another problem.
As far as I understand the question was:
   - Is it possible to use UNIX socket in pure Java (for
RTIA<-->Federate) communication
     which is the default behavior on unix?
     Answer is: NO we can't

Thus the favorite solution is to use 100% java message generated with
the python generator.
Those messages has to be sent on a TCP socket in order to remain 100%
Java (NO JNI).

>
> Sorry for being confused. ;-) Before this email I thought a native Java 
> application, implemented using the Python message generator, is the preferred 
> option.

Again, it IS my favorite option.

As a consequence of the 100% java choice, one have to use TCP socket
instead of Unix socket,
but this is not a real problem because the TCP socket option is
already there, see
RTIA_USE_TCP in the main CMakeLists.txt.



-- 
Erk
Membre de l'April - « promouvoir et défendre le logiciel libre » -
http://www.april.org