[Thread Prev][Thread Next][Thread Index]
Re: larger issue (Re: Linux install misses Tk)
On Thu, Jan 15, 1998 at 03:13:52PM -0500, Platibus wrote:
> Seth,
>
> I am sure there are trade-offs. I think it is execution speed/low level coding
> convenience (perl's advantage) versus maintainability and extensibility (Java's
> advantage). My major reason for using pilotmanager was its availability on the
> sparcstation, and I was quite pleased with it's functionality. I credit this mostly to
> Bharat's ingenuity and not to the fact that he used perl and tk.
A language does not make the program any better. I completely
agree with that. I'm not saying that java is bad (though it might have
sounded that way, I've been fighting with it all day :) ). However, there
is nothing that java would give us that would make it worth rewriting the
system, giving up the vast majority of the work that has already been
done. The system is stable, well written, maintainable and works well
with user supplied modules. There is no reason to rewrite it.
> I don't share your concern about availability of Java on all platforms. Java is going
> mainstream, so stability is just a matter of time, too. But that's just my opinion.
Having painfully dealt with this fact yesterday, I disagree. A
large java system will not run well on many of the non-Sun ports of java,
nor will they exhibit the behaviors that they should. There are many
platforms currently in use that don't have a java port, and might not for
a long, long time. Perhaps they will, eventually, more likely, they will
not.
I know people whom have their cradles hooked up to machines which
are in that category (Alphas running Digital Unix < 4.0 version. No, they
can't upgrade, unless you want to pay for it and port the LISP package
they need over). What should I tell them when you decide to stop
production on this tool and bring it over to java, because java will be
available for everyone and stable some day. Says so right here on the
box. This tool is portable and available for more machines than the java
version will, at least right now. Do you want to give that up? What for?
> I took a quick look at www.gamelan.com and started a search on "perl". There are quite
> a few Java libraries that facilitate perl style regular expressions and the like.
> E.g.:
>
> http://www.cs.umd.edu/users/dfs/java/
> http://www.oroinc.com/downloads/
>
> I guess other people like perl as well and don't want to give up all the perks when
> moving to Java.
REs do not make a good text processing language. There are
intrinsic design decisions in perl that do this, such as list contexts,
tied hashes, etc. You can't build that into java.
> Clearly some low level issues would have to be explored upfront to ensure feasibility,
> though. As far as other conduits go ... I would think that the reverse/re-engineering
> would have to be done with as much automated tool support as possible. Those tools
> could be used to port conduits as well.
My same argument. There is no overwhelming reason to do this.
You haven't given one. The argument that java is more available or more
stable or more portable is untrue. The argument that it is more
maintainable is a matter of opinion. (I'm not sure that an automated tool
can possibly be used to convert idiomatic perl to java. If you can write
this, let me know!)
-Seth
--
"It is by will alone I set my mind in motion"
------------------------------------------------------------------------
***********************************************************
* This is a public mailing list! *
* Please do not publish Sun proprietary information here! *
* - - - - - - - - - - - - - - *
* http://www.moshpit.org/pilotmgr *
***********************************************************