This is so true it hurts. Not long ago I had to rewrite a LDAP interface between a Z/OS mainframe and a SUN Solaris server. The existing code on the mainframe was written in C and it was terrifying. It was what happened when assembler programmers tried to write C. It was full of goto's, not a single line of comments, no indents, and all of the "variables" that were passed in where pointers to pointers of a memory that lived in a mutual operating environment. The C program was invoked by one of four assembler routines, none of which had comments or documentation, and which in turn were all invoked by one of two natural programs. It was a nightmare. I finally reverse engineer the existing code because the prior developer had left, get it running, match up all of the possible return codes, and get that thing perfectly commented, documented, and put into production, where batch jobs start promptly dying. A big meeting was called to figure out what was going on, and it was concluded that my new program wasn't returning a certain value in certain situations. I pull out the original version of the source and say, look, neither is this one. That's when they realize they gave me a version of the source code that was 5 years out of date. So in a couple hours I had to take my perfectly formed program, hack it into pieces and duct tape a stupid little exit to it in order to match that actual production source. Sometimes I think about switching back to academia, but then I remember how much I hate most professors and stay in business.