From: Dave (dfisher_at_pophost.com)
Date: 2002-03-19 12:21:42
On Mon, 18 Mar 2002 16:02:16 -0000, sleeper75se wrote:
>> Somehow my faith was shaken when the only two people who spoke up to
>> say that they would be able to contribute serious amounts of time to
>> the software were Andreas and Dave, and both of them prefer C++. In
>> fact Dave has never even looked at Java.
>
>Hi Jim,
>
>don't worry. If you think it will work, let's use it. I'm not a C++
>fundamentalist, and Java is a nice language - a bit slow perhaps. As
>long as we stay away from the Sun distribution we might even get away
>with it. :-)
I'm not a C++ fundamentalist also, but to decide upon Java right now is
premature. It is the right choice in terms of object-oriented methodologies
since it was expressly created for this purpose (whereas C++ grew out of a
language based on a very different mindset), but it might also be inappropriate
for the kind of applications we are building and the response time we need for
biofeedback.
Also, a quick sidebar. I can't remember who asked about latency in terms of
biofeedback protocols (I think it was Sar?), but I came across a reference
which indicated that a response time of around 200-250ms is considered
acceptable, but these are the "outside" latency limits we would want to shoot
for. Basically, what Chuck has done with ROSHI is amazing with latencies
around 8ms. I think most, if not all, his coding is in assembler, too. That
would account for the phenomenal response time.
So from the moment we pick up that sensor data byte from the device to the
moment that we complete a biofeedback response in our software (image, sound,
etc.), needs to be ~200ms and optimally much *less*.
So this needs to be a serious consideration when thinking in terms of Java, or
even a hybrid solution, such as Java/C++.
>> I really don't know what is the best way forwards, but right now I'm
>> still looking into GCJ. I'm sorry I can't give strong assurances
>> and a clearer lead on all of this.
It will simply take some vertical prototyping to figure out if it can handle
the turnaround speed we need in terms of sensory feedback. If you find that it
is fast enough, then I will have to decide what I need to do -- learn Java or
add my effort to the project in some other way.
>> It is strange how we are such a diverse group, and there
>> is no obvious way to approach all of this. Well, it is
>> obvious if half of the group drop out, but I don't think
>> that counts.
>
>Not sure I follow you ...
What he is referring to is that we have different language backgrounds, and
that to choose Java would say, alientate me, or to choose C++ would alientate
Jim-P. But I trust we will figure that part out.
Dave.
This archive was generated by hypermail 2.1.4 : 2002-07-27 12:28:43 BST