RE: Friendly co-existance

> From: Marc Logemann
> 
> Do you use XML-RPC because its easier to generate the XML when not
> having an API for it? Or is it because of the performance advantage,
> which makes XML-RPC a really useful option in the land of http based
> remote protcols. Are there any APIs for RPG when it comes to XML-RPC
or
> SOAP at all? (sorry, but this is really a home match for java isnt it?
;-
> ))

I prefer clean code.  I hate unused overhead.  Thus, I much prefer
XML-RPC to SOAP.  I consider SOAP to be an excellent example of design
by committee: a bloated set of redundant required pieces with stupid
defaults.  Man, you could tear out 75% of a standard SOAP message and
not miss it.

In fact, I just published a column on web services where I point out
that the overhead for SOAP is just enormous.  Not only that, the code is
so dense that even an expert programmer would have a hard time
deciphering it.  In fact, the folks that designed the API have gone so
far as to say that they don't expect anyone to manually code to it.
That's a pretty scary thought.


> And i just want to make sure that i dont defend every piece in the
java
> world. Lets take EntityBeans in their current manifestation. Its
really
> a painfull spec.

I agree.  There's very little there that I find useful.


> And we totally agree regarding the DB2 for iSeries. I have never seen
a
> more robust database on any platform. This is why i love this machine,
> but i still think that the java implementation could be faster, but i
> have not benchmarked this on various iSeries machines, so its just a
> feeling.

Java on the iSeries is getting better, but it's definitely not at the
same price/performance as other platforms.  But in order to understand
just how difficult it is to even get Java running on the iSeries, you
have to understand the work they had to do in order to get the JVM to
work within the constraints of OS/400.  Unlike other platforms where you
can basically let the JVM go down to the bare metal, with OS/400 you
still have to make sure that you take into account things like object
security.  Then you get little things like EBCDIC/ASCII issues (and code
pages in general) and the fact that the IFS has its own idiosyncracies,
and you start to realize the incredible achievement they actually
managed to pull off.

Blair Wyman and the rest of the folks that built the JVM have done a
fantastic job.  They've made some design decisions I don't agree with;
for example, I'd like to be able to start "system-wide" JVMs that I
could link my job to.  This would not only reduce the startup overhead,
but would also allow jobs yet one more way to inter-communicate - they
could share Java objects!  That would be sooooo cool.

Joe