Re: JGnat support. was NetBeans and ADA?(Co-opt Java?)

[Ingo Marks <adv@region-nord.de>]

Marc A. Criley wrote:

> Wes Groleau wrote:
>> 
>> > However, perhaps one could swipe the Java idea entirely and create an
>> > Ada bytecode compiler with Java support, ....
>> >
>> > We already have the GNAT runtime libraries. Perhaps that would be a
>> > good place to start. ....
>> 
>> If I understand you right, you mean an environment
>> where there would be an "Ada Virtual Machine (AVM)"
>> consisting mainly of the run-time libraries and a
>> byte-code interpreter, and a byte-code that includes
>> primitives for the Ada concurrency features?
>> 
>> Interesting idea.
> 
> I think that's one fork to take, another, though, is simply having a
> rich set of "standardized", portable, and consistent utility packages
> ready to go--comparable to the JDK.
> 
> Marc A. Criley
> Consultant
> Quadrus Corporation
> www.quadruscorp.com

I don't think it's a clever idea just to imitate the JDK.

An Ada based solution must provide far better features than J2EE and .NET 
to make Ada attractive to developers. 

The current big problems Java and .NET suffer is performance. We all know 
that Java is pretty slow compared to C++ and Ada. AFAIK Microsoft has the 
same problem with .NET. Here it's not the C# compiler (they have a good JIT 
compiler for it) but the very slow XML based webservices. In one computer 
journal MS recently recommended NOT to use .NET based webservices. XML is 
based on top of HTTP and that on TCPIP, so there is much overhead.

Ada provides good networking and tasking capabilities. There exist an Ada 
based webserver (AWS) which you can compile your own Ada applications into, 
so that you don't have CGI overhead and compilation overhead (as PHP, Perl, 
MSIL etc. suffer). With Adas tasking capabilities it should be possible to 
write network "task" based web applications instead of page (session) based 
like CGI. Currently you have the choice to use CGI (with C, C++ etc.) or 
script based server modules (mod_php, mod_perl for Apache, for example). 
They are all "paged based", that means they must use sessions to recover 
all user data between page hits (for Java there exists another event based 
solution). My idea is, that in Ada you could use tasks instead of sessions 
or events. For every user a "session task" would be started which surveys 
the user's actions and handle his requests. That way, even realtime 
behaviour would be possible with Ada based web applications. I haven't 
heard about any realtime features in Java oder .NET so far.

Because Ada has standardized compilers (GNAT) for many platforms (Windows, 
Linux, BSD, MacOSX, Sun etc.), AWS could run on several platforms and make 
binary data exchange between different AWS possible. I have played a little 
bit with AWS, it seems not to be pretty stable yet, but it could be 
enhanced.

The advantages of Ada based web applications are evident:

- webserver integrated Ada applications (no CGI overhead, no compilation 
overhead)
- binary data transfer between different AWS platforms (no XML overhead)
- session tasks survey user actions and allow proper actions within given 
time restrictions
- better security (due to Ada features (no buffer overflows etc.))

could result to more powerful and much faster solutions than both J2EE 
and.NET. 

Ingo