Re: [xml-dev] Lessons learned from the XML experiment

["Pete Cordell" <petexmldev@codalogic.com>]

Original Message From: "Uche Ogbuji"
> On Fri, Nov 15, 2013 at 10:56 AM, Pete Cordell 
> <petexmldev@codalogic.com>wrote:
> >  Loose architectural systems, where clearly identified external
> > dependencies are injected in and resolved and dispatched via external
> > configuration are likely to be much more flexible than a system that has 
> > to
> > be crafted with intimate knowledge of a pipeline of how the data should 
> > be
> > processed.
> This might be a matter of how we read what Hans-Juergen wrote, and I admit
> I might have misread it. When he said "integrate schema information from
> 283 different schemas" that to me is not a loose architectural system.  If
> he had said "integrate data from 283 different schemas" then I'd agree.
>
> In other words, for me, the right, loosely-coupled approach approach would
> not be to make a super-schema with all its requisite entanglements from 
> 238
> initial sources, and then use that to direct all the processing.  But
> rather to use pipelines, with some sort of demultiplexer component at the
> input to identify sub-patterns (maybe even at a greater level of
> granularity than the 238 full schemata) and forward to specialized 
> pipeline
> branches for each. It wouldn't magically transform into an easy task, but 
> I
> believe it would be more manageable.
We're definitely on the same page here, perhaps tackling it in different 
ways.  Yesterday we discussed <employer> vs. <com.example.employer>.  One 
way to tackle different types of employer, say private, government, and, 
say, Australian (vs. UK/US etc) might be to make a super encompassing 
<employer> type.  An alternative might be to use <com.example.employer> (the 
one we initially designed our system to handle), 
<com.example.employer.government> and <au.com.example.employer> each 
designed for their own needs, (but hopefully some loose knowledge of each 
other so you can use something akin to duck typing) and dispatch and handle 
them as needed.

But I guess there's many ways to skin a cat!

Pete Cordell
Codalogic Ltd
C++ tools for C++ programmers, http://codalogic.com
Read & write XML in C++, http://www.xml2cpp.com