Jan -

It's always important when discussing SOA to separate
the design principles and best practices from the
various technology options for implementing them.

As we hear a lot on this list, for example, it's
possible to implement an SOA using Java alone.  And if
that's the implementation choice you may not need
loose coupling at all.

The extent to which an SOA is coupled is an orthogonal
question to whether it's an SOA or not.

I believe behind this seemingly endless debate over
custom vs. standard APIs is the age old discussion
about how much information you pack in with the data,
and how much information you place outside the data -
using an interface, a description language, schema
associated with the data, action header, etc.

The reason we probably can't ever get to the bottom of
this debate is that solutions to IT problems are
possible using either approach, or in fact both
approaches.

If you think about old file transfer systems from the
earliest days of computing, we would put everything
into the data that the programs processing it would
need because that was all we had to exchange - and
sometimes only on sequential tape, punched cards, etc.
 

Today we can think about using very simple
asynchronous message queueing systems for all
distributed communications - distributed computing
having the goal of sharing data across computers
somehow.  The question about which method is optimal
cannot be answered in the abstract so the debate
continues without end.  Anyway in the case of using
message queueing for everything, you put into the
message all the information the destination system
needs to process it.  Including a method name and
method signature if what you want to do is invoke an
object.  If you just want to transform to SQL or
something then you can do that instead and you don't
need a method name or input/output argument
definitions.

To me this is a form of standard interfaces that has
been around along time, although the API for it wasn't
really standardized until JMS came along (and even
then that's an API only for the Java "half" of the
world).

File systems such as relative record, sequential, and
indexed sequential have had more or less standard
interfaces for a long time, too.

Anyway, the point is that for me the issue is how much
help to provide the programs processing the data.  If
you are just dumping HTML or XML into a browser then
the browser knows what to do and the system works
great.

However if you are dumping XML into a Java or COBOL
program, especially an existing program in a legacy
application (legacy being basically anything in
production that works), the program is going to need
some help understand what to do.  Custom interfaces
can help - it isn't the only thing, but it is
something that developers understand.

Eric

--- Jan Algermissen <[EMAIL PROTECTED]>
wrote:

> Ron,
> 
> On Feb 26, 2006, at 2:01 PM, Ron Schmelzer wrote:
> 
> > A theoretical exercise in a grand-unified-theory
> of Services serves  
> > very few end-user companies in practicality.
> 
> An essential aspect of SOA is to achieve looser
> coupling - right?  
> This is significant since looser coupling is likely
> to cut down  
> development and development coordination cost, which
> makes loose  
> coupling interesting from a business perspective.
> 
> Some people here (including me) argue that there
> exists a means to  
> achieve even looser coupling than with what current
> SOA proponents  
> suggest. Does that not qualify as a practical issue?
> 
> Or do you suggest that coupling beyond SOA is
> *impossible* and  
> therefore the discussion must be impractical??
> 
> Jan
> 
> 
>
________________________________________________________________________
> 
> _______________
> Jan Algermissen, Consultant & Programmer            
>             
> http://jalgermissen.com
> Tugboat Consulting, 'Applying Web technology to
> enterprise IT'   
> http://www.tugboat.de
> 
> 
> 
> 
> 
> 
> 


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 





 
Yahoo! Groups Links

<*> To visit your group on the web, go to:
    http://groups.yahoo.com/group/service-orientated-architecture/

<*> To unsubscribe from this group, send an email to:
    [EMAIL PROTECTED]

<*> Your use of Yahoo! Groups is subject to:
    http://docs.yahoo.com/info/terms/
 


Reply via email to