8 ms·

I am always amazed by so much effort and so much research in something that is often taken as granted and not really appreciated to its true value by most users

by aphroz 6y ago

I am always amazed by so much effort and so much research in something that is often taken as granted and not really appreciated to its true value by most users (me included).

forgotmypw17 6y ago
An absolutely insane amount of research, testing, and polishing went into Windows 3.11 and Windows 95, and I think the ratio of polish to code reached its peak for Microsoft. Edit: Just to clarify, I am talking mainly of the UI here, not the underlying OS, which was, of course, very basic. Although with virtualized hardware, my experience of running both Windows 95 and Windows ME have been quite good.
- scarface74 6y ago
  Polished? Windows 3.1? It was a shell over DOS. Windows 95 introduced plug and play and a decent driver model. Windows 2000 was peak MS. I could even make an argument for XP. It just didn’t hold up well in the Internet era because of all the security vulnerabilities.
  - forgotmypw17 6y ago
    I'm mainly speaking of the UI, not the guts.
    - scarface74 6y ago
      The UI wasn’t polished either with the idea of the program manager/file manager.
      - forgotmypw17 6y ago
        All I can tell you is that every single UI element was accessible with the keyboard. There is no other desktop environment I've used since then where that was the case.
  - Maakuth 6y ago
    This is common misconception, but in fact since Windows 3.x there was a kernel that did a lot of stuff but made it look like DOS was in control. This was in effect a virtual machine implementation, to make DOS apps behave correctly under multitasking while they were implemented as if owning the whole machine. To facilitate backwards compatibility with DOS drivers, Windows 95 emulated quite a bit of DOS kernel too, with some performance penalty. Raymond Chen's The Old New Thing blog has a number of postings to explain all this, this perhaps being the most informative: https://devblogs.microsoft.com/oldnewthing/20071224-00/?p=24063 https://devblogs.microsoft.com/oldnewthing/20071224-00/?p=24...
    - cptskippy 6y ago
      > This is common misconception, but in fact since Windows 3.x there was a kernel that did a lot of stuff but made it look like DOS was in control. I don't think that's true of Win 3.x, only Win 9x. They are very different beasts.
    - scarface74 6y ago
      That’s Windows 95 not Windows 3.1. Windows 3.1 still used many of the Int21 Dos functions. Windows 95 could use the 16 bit DOS routines for backwards compatibility.
      - Maakuth 6y ago
        I think it is sort of a continuum, with 95 being further on the path begun on Windows 3.x "386 Enhanced mode", or maybe it was like that in Windows/386 already (that was the branding for Windows 2.0). I have a feeling of reading a Raymond Chen blog post about how that worked, but my search skills fail me at this time. However, here's another site summarizing how that mode worked, allowing MS-DOS app multitasking among other things (32-bit disk I/O!): https://networkencyclopedia.com/windows-3-1/ https://networkencyclopedia.com/windows-3-1/
- kanox 6y ago
  Seriously? Windows 95 and 98 were extremely prone to crashes and it took many years for Microsoft to get past a reputation for low quality. It took until Windows XP to get a reasonably stable desktop OS.
  - userbinator 6y ago
    As my other reply here notes, the thin (and permissive) virtualisation layer means that badly written applications could crash the system, but I blame that on the applications more than anything else; too bad that most people seemed to think MS was to blame, when DOS had even less protection. I used 98se as my daily OS until well into the Vista era (at which point I gradually switched into XP), and it was definitely not the "reboot multiple times a day" that some others have experienced, but then again, I also wasn't in the habit of running lots of low-quality applications either. Uptimes of over a month weren't uncommon, and were usually brought to an end more often than not due to myself doing something stupid with the application I was writing. (In fact, I've experienced far more trouble of various sorts with a recent new machine that runs Win10.)
    - iso8859-1 6y ago
      Hmm, using Windows 98 in 2007? I had so many problems with Win98 compared to Win2k, I really have trouble understanding why you'd do this. Especially because you say you only ran high quality applications. High quality software got ported to NT relatively quickly IIRC.
      - jonathanlydall 6y ago
        While I also preferred Windows 2000 once available, most likely he was using Windows 98 SE, I remember it being quite reliable as Windows goes, especially compared to the nightmare that was ME. On install of Windows 98 SE I would always turn off Active Desktop and the fancy looking information left-hand side pane in Windows Explorer. Both used IE to work and would just make everything feel really slow and sluggish. It's also possible that them being disabled improved system stability.
- userbinator 6y ago
  Also, although people often praise the architecture of the NT-based line, I think 3.x and the rest of the DOS-based Windows were far more impressive; while NT is very much a "traditional" OS, 3.x and 9x are in some ways far more similar to a bare-metal hypervisor. This fact becomes very apparent when one tries to write drivers for the latter OSs, wherein you write a user-mode library that basically behaves as if it's accessing the hardware "directly" as an old DOS application would, and then write a "virtual device driver" to virtualise/multiplex the hardware among multiple "virtual machines". There's admittedly far less protection than in the NT world where all applications' (and DOS ones are run in an emulator) access to hardware needs to pass through the kernel, but the thin and permissive virtualisation layer --- where applications can essentially access the hardware directly if not "trapped" by virtual device drivers --- makes for amazingly low resource usage and high performance. A DOS application running in 3.x/9x can interact directly with the hardware and the only overhead is that of V86 mode, comparable to the full x86 vitualisation extensions (AMD-V/VT-x) introduced much later.
mrweasel 6y ago
It's also fascinating how many people back then had almost no background in computer science. Many had al sorts of weird backgrounds, they just found computers interesting and feel into different roles and learn everything they need on their own.

by aphroz 6y ago

I am always amazed by so much effort and so much research in something that is often taken as granted and not really appreciated to its true value by most users (me included).

forgotmypw17 6y ago

An absolutely insane amount of research, testing, and polishing went into Windows 3.11 and Windows 95, and I think the ratio of polish to code reached its peak for Microsoft. Edit: Just to clarify, I am talking mainly of the UI here, not the underlying OS, which was, of course, very basic. Although with virtualized hardware, my experience of running both Windows 95 and Windows ME have been quite good.

scarface74 6y ago
Polished? Windows 3.1? It was a shell over DOS. Windows 95 introduced plug and play and a decent driver model. Windows 2000 was peak MS. I could even make an argument for XP. It just didn’t hold up well in the Internet era because of all the security vulnerabilities.
- forgotmypw17 6y ago
  I'm mainly speaking of the UI, not the guts.
  - scarface74 6y ago
    The UI wasn’t polished either with the idea of the program manager/file manager.
    - forgotmypw17 6y ago
      All I can tell you is that every single UI element was accessible with the keyboard. There is no other desktop environment I've used since then where that was the case.
- Maakuth 6y ago
  This is common misconception, but in fact since Windows 3.x there was a kernel that did a lot of stuff but made it look like DOS was in control. This was in effect a virtual machine implementation, to make DOS apps behave correctly under multitasking while they were implemented as if owning the whole machine. To facilitate backwards compatibility with DOS drivers, Windows 95 emulated quite a bit of DOS kernel too, with some performance penalty. Raymond Chen's The Old New Thing blog has a number of postings to explain all this, this perhaps being the most informative: https://devblogs.microsoft.com/oldnewthing/20071224-00/?p=24063 https://devblogs.microsoft.com/oldnewthing/20071224-00/?p=24...
  - cptskippy 6y ago
    > This is common misconception, but in fact since Windows 3.x there was a kernel that did a lot of stuff but made it look like DOS was in control. I don't think that's true of Win 3.x, only Win 9x. They are very different beasts.
  - scarface74 6y ago
    That’s Windows 95 not Windows 3.1. Windows 3.1 still used many of the Int21 Dos functions. Windows 95 could use the 16 bit DOS routines for backwards compatibility.
    - Maakuth 6y ago
      I think it is sort of a continuum, with 95 being further on the path begun on Windows 3.x "386 Enhanced mode", or maybe it was like that in Windows/386 already (that was the branding for Windows 2.0). I have a feeling of reading a Raymond Chen blog post about how that worked, but my search skills fail me at this time. However, here's another site summarizing how that mode worked, allowing MS-DOS app multitasking among other things (32-bit disk I/O!): https://networkencyclopedia.com/windows-3-1/ https://networkencyclopedia.com/windows-3-1/
kanox 6y ago
Seriously? Windows 95 and 98 were extremely prone to crashes and it took many years for Microsoft to get past a reputation for low quality. It took until Windows XP to get a reasonably stable desktop OS.
- userbinator 6y ago
  As my other reply here notes, the thin (and permissive) virtualisation layer means that badly written applications could crash the system, but I blame that on the applications more than anything else; too bad that most people seemed to think MS was to blame, when DOS had even less protection. I used 98se as my daily OS until well into the Vista era (at which point I gradually switched into XP), and it was definitely not the "reboot multiple times a day" that some others have experienced, but then again, I also wasn't in the habit of running lots of low-quality applications either. Uptimes of over a month weren't uncommon, and were usually brought to an end more often than not due to myself doing something stupid with the application I was writing. (In fact, I've experienced far more trouble of various sorts with a recent new machine that runs Win10.)
  - iso8859-1 6y ago
    Hmm, using Windows 98 in 2007? I had so many problems with Win98 compared to Win2k, I really have trouble understanding why you'd do this. Especially because you say you only ran high quality applications. High quality software got ported to NT relatively quickly IIRC.
    - jonathanlydall 6y ago
      While I also preferred Windows 2000 once available, most likely he was using Windows 98 SE, I remember it being quite reliable as Windows goes, especially compared to the nightmare that was ME. On install of Windows 98 SE I would always turn off Active Desktop and the fancy looking information left-hand side pane in Windows Explorer. Both used IE to work and would just make everything feel really slow and sluggish. It's also possible that them being disabled improved system stability.
userbinator 6y ago
Also, although people often praise the architecture of the NT-based line, I think 3.x and the rest of the DOS-based Windows were far more impressive; while NT is very much a "traditional" OS, 3.x and 9x are in some ways far more similar to a bare-metal hypervisor. This fact becomes very apparent when one tries to write drivers for the latter OSs, wherein you write a user-mode library that basically behaves as if it's accessing the hardware "directly" as an old DOS application would, and then write a "virtual device driver" to virtualise/multiplex the hardware among multiple "virtual machines". There's admittedly far less protection than in the NT world where all applications' (and DOS ones are run in an emulator) access to hardware needs to pass through the kernel, but the thin and permissive virtualisation layer --- where applications can essentially access the hardware directly if not "trapped" by virtual device drivers --- makes for amazingly low resource usage and high performance. A DOS application running in 3.x/9x can interact directly with the hardware and the only overhead is that of V86 mode, comparable to the full x86 vitualisation extensions (AMD-V/VT-x) introduced much later.

mrweasel 6y ago

It's also fascinating how many people back then had almost no background in computer science. Many had al sorts of weird backgrounds, they just found computers interesting and feel into different roles and learn everything they need on their own.