winscape 2258

I think they achieved fundamental brokenness by accident- by trying to be different and innovative, but not making performance and quality and utility of primary importance, they succeeded in making things look different but not making them work better- or even as well in many cases. I think quality has never mattered much to Microsoft- they've always been better at sales than software.

I think in large part it was unconfusing myself about x86 which is a difficult architecture to learn on & using DEBUG makes it worse. I came from a 6502 background which is quite approachable OTOH. If I'd moved to MIPS or something along those lines instead of x86 I think I wouldn't have had as much trouble.

One aspect of the problem is DEBUG's x86 syntax is archaic beyond belief and since it offers the programmer no help whatsoever and the only programming reference you tend to have is the DOS int 21h api, a greatly mistaken understanding about how programs and OS's are structured readily develops. It took getting a macro buttembler and symbolic debugger and integrating asm wC to really start to unlearn. The best x86 buttembler I used was Borland's tasm. They implemented an enhanced syntax "Ideal Mode" you could put the buttembler in which sanity checked and formalized the asm syntax so it made much more sense and rejected a number of bad programming habits. Or you could leave it in the "broken-butt confusing" syntax mode for backwards compatibility.

I've not used DDT so can't comment in detail. However Solaris's dtrace seems to do the kinds of things you talk about DDT doing. Probably various bits and pieces of DDT like functions been implemented over the years to help with kernel debugging but naturally vary with the OS & version & architecture....

Gregm