Re: Porting SWARM to JAVA or C++

[Christoph Schlumpf]

Manor Askenazi writes:
>Christoph Schlumpf writes:
>
>  ... It should be no great problem to do swarm in java (besides of
>  manpower and time) ...
>
>This discussion is obviously academic (which sort of makes sense on
>a list with so many .edu addresses), nevertheless there comes a point
>where ignoring real world constraints smacks of silliness not academia.
>
>In other words, yeah, Christoph, if you can find the manpower and you
>can find the time, please, increase our options. Right now, we are
>still using funding that we *miraculously* got from DARPA to finish
>this *FREE* version of Swarm which started at a time when Java didn't
>even exist! Also, keep in mind that we've had maybe 1.5 programmers
>working on this at any given time - so you see that manpower, time
>and, by implication, money, are for some obscure reason, less than
>trivial constraints...
>
>Having said that, the real reasons for remaining in Objective-C are
>and will remain technical, and for an explanation of these issue I
>defer to Roger's reply to a similar question which you will find in
>our hyper-mail archive:
>
>http://www.santafe.edu/projects/swarm/archive/list-archive.9607/0001.html
>
>The discussion is slightly dated in that it disregards the "Reflections"
>API, which you mentioned in your mail and which presumably did not exist
>when Roger wrote his mail. However, issues like the lack of pointers and
>low-level data structures remain, and seem, IMHO, to constitute serious
>drawbacks when considering a port to Java.
>
>Regards,
>
>Manor.


Dear Manor,

I never intended to attack you and the whole swarm-team. If I stated that
it is no problem to do swarm in java I meant (as I stated earlier in that
email) that there are no technical reasons why one can't build a comparable
powerful messaging-mechanism in java as it has been done for swarm in ObjC.
I worked for quite some years in a small software firm. So I wasn't aware
that there are people around in this group who don't know how much time and
money is in such a software. I put this in brackets not because it is not
essential but because it is obvious that time and money is always a
constraint to doing a software-projects.

I think ObjC was really the right choice the time the swarm project was
started. But as you write, a java port could 'increase our options'. I
never stated that I'll do such a port or that anyone has to do it. I just
mentioned that I'd like to have such a port and that, with its new
abilities, it should be possible to do it in Java.

Besides of the Reflections API Java always had pointers (but no pointer
arithmetics). There is much money going into the development of integrated
visual development environments for Java. While there is perhaps too much
hype around java there are obvious reasons which makes java a considerable
OO language. The main disadvantage of JAVA for swarm is that it doesn't
allow for direct memory access and management. Even in compiled form a java
based swarm will eventually be considerably slower if you work with large
models on multiple processors. But with large models Java helps to prevent
you from severe bugs by its Typesafty mechanism and by its automatic memory
management.

I'll like to work with swarm in ObjC in the near and far future. You are
doing a great job and everything else is just vaporeware.

Regards,

Christoph.