people were told after a three day course with von neumann (herzfeld) or a weekend skitrip - (i forgot who, it's one of the famous guys in hiltzik's book), or one week training "you could program a computer" (alan kay) - or you knew enough about computing to take the job.

"now with 1000 professors and more graduates" as charles herzfeld says, it's harder to have a nobel idea as it is harder to find a good programmer, as joel complains in his story about lazy kids.

online course for a test.

what happened to the idea training on the job nowadays?

you know my opinion on this. joel knows that, what he does, is conservative envious bullshitting. he does not pose the question in a serious way but in a superficial way, finding the responsibility for existing phenomenes in modern easified programming languages and the teaching of them.

this is a critique of a certain kind of progress in computing(ultimately Smalltalk, its interface and the ideas behind it) which aimed bringing the top layers of application programming nearer to their respective real world domains and never aimed to write infrastructure SW. Whoever is doing basic OS or any infrastructure (like virtual machines, optimizing compilers, process communication) in Java uses it to simulate features of those wares, not serious implementation, that's done by an ever optimized compiler or virtual machine, written in C. those c and apl and assembly and lisp gurus did some good and so much bad sw that luckily everyone has forgotten about. the real question is a deeper cultural one.

why is the mathematical and scientific base for good programming going away. the top brass in those fields is still very good. but the average is down, isn't it. Why then? There must be many factors, i guess, not the least that kids have a very different set of complicated requirements to master nowadays. No one can be afford to be a geeky uncommunicative freak anymore like most of those sweet scheme programmers and guys able to express recursion in 360/assembly were. their wives and kids and friends just don't allow that anymore. To name just 1 factor.

You also know my base theorem: electricity and the way it connects everybody destroys straight forward literalization (which was sustained by communication forms like letters books and geometry), deliteralization drastically lowers the average (important, not maximum or so) mathematical ability, lower average mathematical abilities lower the average quality in all fields that require different sorts of calculus / formalization / high level abstraction ... (algorithm, recursion, field structure...). It definitely is not Excel (think about MBAs that can do addition and multiplication and already fail on division), pocket calculators, Basic and Java that did that to us. It is also not the laziness of modern kids. But ok, want more good C programmers, more cool usage of recursion, complex set data manipulation, Latin and Greek abilities? Beat up all of them kids, discipline all of them into those abstract formalized systems and you'll get got more people apt for C and Lisp, cause that was, how it was done the last time. And some still where happy, cause you could be talented, have nice parents, teachers and priests and still swim with that culture. We don't know yet how to do it in a better way and that's what Joel should write about.

Noch eine Bemerkung: Joel linkt Cringely, der demonstriert warum Google besser ist. Sie sind besser, aber er vergleicht Äpfel mit Birnen. MS ist immer bei den PCs und bei Userfantasien, so wie IB; immer big iron war, large scale business and scientific computing war. Nie waren und sind die bei der Netzinfrastruktur, oder halt nur als afterthought. Dort ist Google, und die sind besser, algorithmisch und überhaupt. Aber weit weg von den Usern, außer mit ihren minimalistischen Interfaces. MS-Software brauchte nie viel Wissenschaft. Das sind einfach nur mühsame Features und die Algos dazu sind alle schon seit 30 Jahren publiziert. Samt Optimierung. MS is about selecting features in a PC, Google is about infrastructure that practically none of their users understand. Google is the ultimate go for techie domination. A lot more than MS ever was. Cool. And Google is back to the 60ies and to a holy Priesthood of Computing.

Given all these circumstances it seems to me the real problem is neither java schools nor kid laziness (which is there as my daughter who is doing electrical engineering and tells me there are so many kids that even do not know how to open a shell). My feeling is that large gaps are opening in the long chain from top science and its application over to the users of the outcome of it. Lets look at those gaps! And we don't need better tools for digesting this changes but new and better ways of scientific intereducation.

John von Neumann ist tot.