Re: [xml-dev] RE: Declarative programming requires a different mindset

[Stephen Green <stephengreenubl@gmail.com>]

Isn't it the case that there are declarative languages where
the outcomes of computer programs are already very well
defined and merely need some guiding or, as is the usual
term, 'configuring'. As with Postscript - the requirement is
already clear that the input is sent to a printer to be printed
as the output and there only remains, in most cases, the
extra formatting: In fact the printing programs have a primary
task which is to bring about the printing of what they are sent
which in some cases they could reasonably do without any
further instructions at all. So in such cases where there are
clear requirements described in the spec there is less for the
programmer to do. So declarative languages have most of the
program requirements in a specification and in the system
which conforms to the specification and there is little need for
programming, just for what amounts to a kind of configuration.
The program is already written; all it needs in addition to the
input (or as part of the input) are some extra instructions which
can be in the form of a 'declaration' of the extra requirements
(in some cases, extra to a default behaviour or output).

I think that has always been a goal of computer development
and it is fulfilled especially well in our times due to the invention
and widespread adoption of XML which makes 'configuration'
work so very nicely. So nicely, I think, that 'configuration'
languages can take on a lot more of the detail and power so
that they start to look less and less distinguishable from
computer programs (hence this debate). So much the case that
people formerly involved in writing programs now find themselves
confiuring them declaratively. E.g. XSLT declarative stylesheets
can do so much because the XSLT processor and the
expressiveness of the XSLT language are so powerful that the
stylesheet is practically 'king'. e.g Saxon does the heavy lifting so
that the stylesheet gets to drive without having to work too hard.
This sometimes results in the declarative part of the system (XSLT)
becoming Turing-complete as much as the programming language
used to write the processor of the declarative language (C++, etc).

---
Stephen D Green