[Rxtx] rxtx moving from JNI to JNA (was Re: About JRE crashes)

[George H]

On Wed, Aug 18, 2010 at 11:52 AM, Kustaa Nyholm
<Kustaa.Nyholm at planmeca.com> wrote:
>
>> Instead of trying to convince everyone how easy JNA is based on a simple
>> printf example in a Wiki - why don't you prove it?
>> Implement RXTX using JNA and make us all believers.
>
> I have no need to convince you or anyone else, I'm happy with rxtx the
> way it is at the moment. As long as I can get a pre-compiled binary that
> just works for the platforms that I need it for, it is all that I need.
>
> I was talking in the context of someone's suggestion that a re-write is
> contemplated and was merely suggesting that in that case it might make
> sense to to get rid of the C stuff and the associated practical problems.
>

Yeah it seems that context got taken out of hand.
Adrian Crum was talking about re-write here
http://mailman.qbang.org/pipermail/rxtx/2010-August/6848586.html

There is still the part of performance issue mentioned by many. And
this one got lost from the thread since the subject changed
http://mailman.qbang.org/pipermail/rxtx/2010-August/6848590.html

If you are happy with how RXTX is then continue to be happy, provide
input and let developers be open to experimentation. It seems you are
pretty adamant about using JNA. It seems to be good for a lot of
things but maybe not for RXTX.

The only way i can see JNA being used productively in RXTX is to be
the proxy pattern between the Java code and the Native rxtx code. From
what I read on this list and from looking at the native source code,
it is likely that a performance hit will be a big issue if the code is
moved up the stack to java. Should some people disagree then they
should do a performance test with a test lib where they shifted some
of the code up the stack.

>> Until then, you're just making things up.
>
> I don't think that was a gentlemanly remark worth of a reply.
>

Don't take it personally, we are all computer people, we trust
evidence based on experiments and benchmarks more than people's
claims.

--
George H
george.dma at gmail.com