This is kinda like judging C by how bad Algol was. Fortran 90 and its successors should be looked as a separate language from the old fixed format Fortrans, and they are the currently IMO the best we got for supercomputing purposes (i.e. I'd prefer it over C/C++, in that specific domain). That being said, I hope Julia pushes through as a replacement. Note that I'd never advocate for Fortran in any capacity outside of that domain, e.g. for system programming.