Posted-Date: Sat, 4 Jan 1997 22:55:32 -0600 (CST) Message-Id: <199701050455.WAA26755@mail.texoma.net> From: "Mark S. Teel" To: "Benjamin D Chambers" Cc: "DJGPP" Subject: Re: 32-bit filenames Date: Sat, 4 Jan 1997 23:00:24 -0600 MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit ---------- From: Benjamin D Chambers To: mteel AT texoma DOT net Subject: Re: 32-bit filenames Date: Saturday, January 04, 1997 12:26 PM On Sat, 4 Jan 1997 09:52:21 -0600 "Mark S. Teel" writes: > >I'm not sure what we are debating here now, but, it is more than >register >size. It also involves CPU bus width (16 vs. 32 bit) and the fact Well, the 32bit bus was introduced with the 386 as well. >that the >"old" 16 bit intel cpus required the segment-offset approach to >address >memory outside of 1 meg. This approach (and IBM's That was simply to maintain compatability with older processors. The chips also supported flat memory models (starting with the 386), so the point is moot. initial choice of >Intel >vs. Motorola 68000 for thier first PCs) has confused/restricted/led to >our >ridiculous dependance on Microsoft for sub-par software for the last Our rediculous dependance on M$ has been because they got M$-DOG shipped with machines. (I think stores should ship PC's with Linux :) >17 >years or so... Those of us who have programmed 68000s and flat memory >model cpus in general have disliked Intel/segment-offset/Microsoft/PC >programming for years! Hail djgpp and GNU! I agree that flat memory is the best (You wouldn't believe how happy I was to find that allocating 4megs of memory all at one shot actually _WORKED_ in Linux - and when I found DJGPP, I was even happier :) The point I was arguing was that Tom Wheeley said the 16bit/32bit thing had nothing to do with registers - when in fact it's nothing BUT the register size :) ...Chambers Well you are partly right - some of the first 386s had 32 bit registers and 16 bit buses (remember "SX"). I thought this had something to do with "16 bit filenames" or some such rot - the 8.3 DOS file format had nothing to do with register size, just another DOSism we have lived with. And the fact that 386s and up supported a flat memory model does not relieve my angst in that the whole architecture is very clumsy when used in this way and the backward compatibility is exactly what continues to make Intel CPUs so distasteful to me. The filename restriction was caused by the same forward thinking that gave us the 1 meg restriction: "no one will EVER need longer filenames than that (or more than 1 meg of addressable space)". We continue to live down these initial decisions. When is PowerPC going to take over the PC CPU market? I'm waiting... MST