Re: [xml-dev] Build applications using the "simplicity stack"

[Thomas Passin <list1@tompassin.net>]

On 4/1/2014 5:46 PM, Arjun Ray wrote:
> On Tue, 01 Apr 2014 04:03:06 +0100, Ihe Onwuka <ihe.onwuka@gmail.com>
> wrote:
>
> | If you look at the industry the actuality of a simplicity stack for
> | most is using (or being told to use) the subset of tools you already
> | understand (appropriate or not) to solve every problem you encounter.
>
> Actually, this is not at all facetious.  The impact and influence of
> software one has become accustomed to, on habits, thought processes -
> and ultimately, even design decisions - is vastly underestimated, as
> is the propensity to follow paths of least resistance while believing,
> quite sincerely, that one is better than that.
Yes, indeed.  And working with the tools that you know is often the most 
effective way to approach a problem.

I like to use XML and especially XSLT when they fit - because of the 
flexibility they can give - but it depends on how large and complex the 
task is.

I have a personal browser-based application that uses topic maps to 
model my browser bookmarks.  I get input from a netscape-format HTML 
browser file, from Internet Explorer bookmark files (each a separate 
file in itself), from various files on my computer, and possibly from 
XTM (an XML language for topic maps).  I need to transform all of these 
inputs into a pre-JSON javascript file.  So I use a range of languages 
that were comfortable and easy for me to write:

- Everything is choreographed by a shell script;
- I use Python programs to convert this netscape HTML bookmarks and the 
IE bookmarks to an intermediate xml format.
- I walk certain directory trees and form appropriate URLs using a 
python program.  The output is the intermediate XML format.
- I use a short XML driver file to specify which XML files to process in 
the next step.
- I use an XSLT file reading the ddriver file to combine and transform 
all the intermediate XML files to a single javascript output file.

The browser application loads the javascript file along with the topic 
maps engine.  From time to time I run the shell script to update to my 
current bookmarks.

So I'm using shell script, python, xml, xslt, and javascript.  Each step 
could easily be done with some other language, but these work very well 
for me. [I'm actually still using XSLT v1.0, and an old, old version of 
Saxon (thanks, Michael Kay!).  If it's working, it's golden.  And it's 
been working for ten years].

I think this stack is typical.  For someone else doing a different job 
it might be perl or awk instead of python, maybe lisp or haskell instead 
of xslt, json instead of a home-grown complete javascript file, or a 
mixture of java and jython.

OTOH, to turn a system like this over to someone else for production and 
maintenance could be a whole different story!

TomP