Tim Bray says that high amounts of COBOL code in production is "scary". I disagree. Well written COBOL code running on a mainframe is rock solid. The code can run for decades, partly why we had the Y2K problem to begin with. To compliment the stability of the language, it is mostly in use on ultra-stable hardware: mainframe computers. Who would want to give up the stability of a mainframe to move to a J2EE application running on a windows server? Not I said the little man. Imagine missing a payroll check because a new windows exploit took out the server on payday. That isn't to say that mainframes don't have to be patched, but it seems like, at least traditionally, the patch rate is much much lower than other platforms.
Secondly, COBOL is a fine language for Reporting and Records manipulation. Definitely one of the best report-generating languages (if not
the best); dare I say it's better than Perl at record manipulation and reports? Regex gives Perl a definite edge, but I think the comparison is too close to call. I for one would *kill* (not literally) to have COBOL style record handling as an option in C/C++/Perl - heck, even Java should have it if it doesn't already (I'm not a regular user of Java and certainly no guru with what features Java offers). Many modern languages offer some record manipulating features, but IMHO none are quite as suave as COBOL's - but then with everything drifting towards databases, perhaps we could make the argument that the flat-file is dead?
