[Help-glpk] Re: XML Result Output

[Andrew Makhorin]

>A problem all solvers face is the non-uniform output from the solver.
>We are constantly faced with reading a text file, that is not always the
>same depending on feasibility, and trying to extract the results.  I've
>been amazed that nobody has built an XML output from any commercial or
>open source solver so standard XML parsing routines are available for
>extracting and recording the results.  I would like to see an XML output
>option to aid in application integration.

I am not well familiar with the xml technique, so I read some papers
about that, in particular, some papers (pointed out by Robert Fourer)
about using xml for coding optimization problem instances. Must say that
I do not understand why xml is better than plain text (for example,
dimacs format used for coding some combinatorial problems). Except
parsing all other problems persist. In any case there must be someone
who must develop all neccesary data structures, because xml itself has
no specific data structures. But once the data structures have been
developed, coding them in any suitable format is not a problem at all.
In my opinion the lexical set as well as some syntactic conventions
used in the C language is much more suitable for coding complex data
structures than xml. Could anyone explain me in a few words where I am
wrong? Let, for example, I need to develop a format for coding, say,
data for the minimal covering set problem. What I could gain coding the
data in xml format rather than in plain text format (like dimacs)?

- Andrew Makhorin

ps. There is no doubt that some *communication* formats must be
introduced in glpk to allow interchanging data between glpk and the
user's application programs. Unfortunately I never met any more or
less standard formats (except mps) suitable for this purpose.