OT: was: Programmer's conundrums

Tony Duell wrote: Tony, I sat on this for a few days since I try not to take the bait, but I finally couldn't resist. If it was just this one comment, I'd let it go, but you frequently post messages dripping with disdain like this one. As an engineer with more than twenty years of experience, I found what you said insulting and misinformed. When I started in 1985 when a 6 MHz 68020 Sun machine was hot stuff, CPU cycles weren't really all that free (although the old timers on the list will laugh at that). Now I work at a company with a farm of 5000+ CPUs running at 3 GHz, each with 4 or 8 GB of DRAM each (more for the subset set aside for large jobs). There were crap engineers when I started and there are crap engineers now. I'm sure there were crap engineers well before my time and will be after I retire. The flip side is there are very many very smart people still in the trenches creating new work. To use a very broad brush and claim engineers today are inferior for making rational decisions regarding resource allocation is just misguided smugness. Can you imagine routinely designing 50 M gate chips in 12 months without having significant CPU resources to aid in the process? The heroic engineering of yesteryear that you so admire, and rightly so, is no longer applicable. The constraints have changed, and so have the methods. Attempting to lay out a modern chip by hand would produce something riddled with errors; it would take forever to complete, and in the end it would be much larger than the one using CAD tools. Relying only on hand checking for programs of real complexity is impractical. I suppose every time I change my verilog code I could spend a day or a week contemplating it to foresee the unforeseeable, and after I got fired for being unproductive, I'd have a lot more time to do things that way. No, it is far better to do the best you can in a reasonable amount of time and let a dumb and fast pile of CPUs run regressions after any changes. At more than one company I've worked for it is policy to run a small regression after any change even if the only thing that has been changed is a comment. It is important to make a distinction between science and engineering. Scientists try to compute things exactly as possible; engineering is more about having the wisdom and experience to know that although it is possible to compute something to the 15th digit, for the given application, 6 digits, say, is enough. Engineering fundamentally is about cutting corners; engineering is pragmatic. Here is a thought experiment. If the designers of the <insert name of some machine you admire> had access to the resources of today with the constraints of today, do you think they'd still go about it the same way?