Hello David,
I wanted to get your insight on some of these issues ...
> Today a JAVA application might run 5 to 100x slower than an
> application written in C.
I agree ... to a point. First of all, this is a very broad statement that doesn't truly involve all of the various issues. I'm guessing that you are comparing identical platforms, and comparable skills of programming, and that the platforms are Windows or UNIX? (Also, I'm not completely blinded that the success of Java is there ... I'm a big fan of Limbo from Lucent ... I think it's a much more elegant solution! And it runs Java!)
In order to measure the *true* benefits of Java you also need to take into account the maintainability, cross-platform benefits, and the overall cost benefits that make software more reuseable across a broader market. What about development time and cost to develop that C application? And how monolithic is the result? This is a strategic, business issue ... not just who runs faster ... the question is "Is it good enough?" or "It is useable and cost effective?" ... is it fast enough to be an effective investment?
I tend to sell customers on the *options* that Java provides them ... the freedom to choose in the future! Long term investment in IS assets.
My question to you is:
1. Do you think that Java will always be 'slow' and that Intel and
others will not work to resolve this?
2. What kind of time are we talking? A year?
I don't know ...
> Writing NLMs are very ugly, when compared to JAVA.
Uh ... as an NLM developer, and a Java developer I have to disagree! The Java tools are pretty bad these days ... and Java in no way lends itself to the development of high-performance services. Not today ...
> NLMs really don't support a GUI, and there are only a handful of
> tools out there to do it.
And the reason for a GUI on a service is? Why would I ever want to burden a high-performance I/O service with a GUI? GUIs should be developed independent of the service, and through a mechanism that provides security, access control, and network transport independence ... I would argue for a directory interface, unless real time information is needed in which case SNMP is a better standard solution ... both of which NetWare fully supports.
> You put 1,000 C programmers in a room, and I would bet you wouldn't
> find a single one who has ever written a NLM, compared to maybe
> half which have piddled, or are now fluent in JAVA.
And from my basic mathematics I remember that large numbers of people tend to cluster around the center of the bell curve ... that's average! The people who push the limits will end up at one end or the other. At my last couple of start-ups, we actually made a lot of money, quite easily, just because not many people have the expertise and the skill to right NLMs.
A Windows app? Easy ... I can write crap code and if it blows up, the user will just re-boot ... they're all trained by Microsoft! ;-)
> NOVL certainly has the right idea here. Unfortunately, they don't
> control their destiny. JAVA today is just too slow and bulky. Corel
> learned this the hard way with their wordperfect office written in
> Java.
I agree ... ;-)
> David
Scott C. Lemon |
