From: muta...@gmail.com   
      
   On Friday, July 23, 2021 at 12:04:38 PM UTC+10, anti...@math.uni.wroc.pl wrote:   
      
   > > > Another things: you write as you invented soemthing. Various   
   > > > segmentation schemes were studied and in 1985 it was well   
   > > > known that technically you could use different segment   
   > > > shift. But it was also known that such machine would   
   > > > have no advantages, so nobody tried to make it.   
   > >   
   > > Ok, so just for the record.   
   > >   
   > > It was well-known by 1985 that at least SOME 8086 programs   
   > > (in all memory models, even huge) could be run on the 80386   
   > > and have a 512 MiB address space, but absolutely no-one saw   
   > > any advantage to 512 MiB vs 1 MiB and would prefer that   
   > > ALL their 8086 programs be restricted to 1 MiB for a number   
   > > of years and then completely fail.   
   >   
   > Sorry, you can not repeat correctly. It was well-known that   
   > you could run some 8086 programs on 286 and 386 and use more   
   > memory (up to 16 M on 286, up to 512 M on 80386). It was   
   > also well understood that arbitrary segments origins on 286   
   > and 386 gave more possibilities than your "segment shift",   
   > so nobody was intereded in machine using shift bigger than 4.   
      
   Ok, my interest is in "advice for 8086 programmers" to be   
   published in 1978. So the "segment shift" concept will   
   already exist by the time the 80386 is produced.   
      
   > Concerning using more memory: it was known what to do and   
   > there were ready made solutions (DOS extenders). AFAICS   
   > there was little interest in making "universal binaries",   
   > that is one which worked on 8086 in 1M and on better   
   > processors using more memory.   
      
   I was always interested in doing that. I have a fundamental   
   expectation that when I plug in an 80386 with 4 MiB of   
   memory that all my MSDOS programs start using all of   
   that extra memory, with no change to any software. Just   
   replace a 1 MiB 8086 with a 4 MiB 80386 and that's it.   
      
   Oh, it's OK if I have to buy MSDOS-32 for that to work though.   
      
   If my expectation above is "bizarre", so be it. I'm a bizarre   
   person and proud of it.   
      
   > Concerning "no one": there was a lot of various developements,   
   > most is now forgotten. It is possible (even resonably likely)   
   > that some obscure programs tried to use more memory and   
   > be compatible with 8086. But one can be resonably sure   
   > that nobody wrote programs targeting your PDOS.   
      
   They only needed to target MSDOS, with some very slight   
   adjustments to 16-bit MSDOS, to prepare for the future.   
      
   > And   
   > using 512 M in program capable of running on 8086 with   
   > 1 M was in little (maybe no demand) simply because in 1990   
   > machine with 512 M memory would be extremaly expensive   
   > special-purpose machine, and later in general ability to   
   > run on 8086 was of limited use.   
      
   512 MiB is just the maximum. The principle applies to a   
   2 MiB machine. I believe 2 MiB machines existed in 1990.   
      
   > Little correction and addition: there was period when Windows   
   > supported 8086, 286 and 386. Windows applications were   
   > supposed to run on all those machine. So there is body of   
   > application that run on 8086 with ability to use more memory.   
   > You prefer DOS, so probably this is of no iterest to you.   
   > But this shows that it was known how to do this and it got   
   > some use. OTOH I never saw Windows running on 8086.   
   > I saw guy running Windows on 1 M 286. Guy was satisfied   
   > that it run at all, my opinion was that it was unusably   
   > slow. I remember that when Miscrisoft annouced that   
   > Windows no longer supports 8086, there was a comment   
   > that progrmmers are very satisfied, trying to make   
   > programs run on 8086 was pain with little reward.   
      
   I am not competing with GUIs. I am only interested in   
   text mode OSes and applications.   
      
   > Note that at this time most programs were commercial. Simple   
   > business rule is that people who have more money can pay   
   > more, so you should set higher price for them. But how   
   > to recognize folks with more money? One simple trick is   
   > to look what processor they have, if it is 386 than   
   > owner spend notrivial money on processor/computer so   
   > can pay more for software. So software house would have   
   > "low end" 8086 version and more expensive 386 version.   
   > 386 version may be faster due to 32-bit instructions   
   > and there is more place for features but it would not   
   > run on 8086. Put it differently: why work hard and   
   > get less money, when you can work less and earn more?   
      
   I am not trying to make money. I'm trying to develop   
   clean text mode applications for the 8086.   
      
   > Anyway, you have your own desires, but most people moved   
   > on and are happy to consider 8086 just as history, not   
   > as something to be a serious developement target.   
      
   I'm not claiming nor do I desire to be "most people".   
      
   Currently I am working with some university-aged Filipinos   
   who seem to be having a barrel of fun running bwbasic   
   that runs under both Windows 10 and PDOS/386.   
      
   My wife had predicted that they would have no interest in   
   doing that because the younger generation were only interested   
   in smartphones.   
      
   Someone, somewhere, got some seriously incorrect   
   information.   
      
   Who would have thought that even modern people were   
   interested in learning how computers work, instead of spending   
   their whole lives making Tik-Tok videos?   
      
   BFN. Paul.   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)