22

According to The launch of Windows 3.1 in Japan was such a big event that they even called it “Windows Day” (Bullfrag):

When Microsoft released Windows 3.0 in the United States in 1990, they were selling about a million copies a month. Meanwhile, in Japan, that version had a terrible reputation for being buggy and sold less than half a million copies in two years. This changed significantly with the release of Windows 3.1.

Why exactly did the Japanese version of Windows 3.0 not work well?

It worked well enough for Americans, so presumably the problem was with the Japanese language support.

According to Unicode in Microsoft Windows (Wikipedia):

Windows NT was the first operating system that used "wide characters" in system calls.

At that time, this was widely considered to be the canonical way of 'supporting Unicode'.

So Windows 3.0 didn't use wide characters. On the other hand, it must have done something about Japanese-language support; Microsoft could not possibly have tried to sell an ASCII-only operating system in Japan in 1990.

What exactly did it use, and why didn't it work?

13
  • 14
    Shift-JIS, probably. But maybe encoding was not the reason why Windows 3.0 was not big in Japan; maybe Americans just have more tolerance for substandard products? Mar 4 at 18:52
  • 9
    3.0 wasn’t particularly good, 3.1 was better.
    – Jon Custer
    Mar 4 at 18:56
  • 11
    @JonCuster: That reminds me of a nice riddle: "What's the difference between 3.11 and 3.1? In the Windows 3.11 calculator, 0.01. In the Windows 3.1 calculator, 3.11-3.1 will yield 0.00.
    – supercat
    Mar 4 at 19:36
  • 19
    I don't remember Windows 3.0 being a success anywhere in the world. Win 3.0 was considered quite slow and extreme buggy with no real advantage 8btu did support charsets). 3.1 in contrast brought several essential improvements, like Drag&Drop, cut&paste using CTRL-C/X/V, Win32s subsystem for 32 bit programs, registry, arbitrary resolutions with up to 32 bit colour, native network support and many more. I can't talk for the US, but in Europe, Win only became a platform to be recognized with 3.1 as system for 3/486 class machines supporting 32 Bit programs - with 3.11 being the real turn point.
    – Raffzahn
    Mar 4 at 19:51
  • 5
    @Raffzahn Win 3.0 only seems like a product without much impact because it is shadowed by the even far greater success of Win 3.1 which came quite quickly after. As the question text says, Win 3.0 was selling a million copies a month in the US not long after release. Not exactly a fringe product.
    – RETRAC
    Mar 4 at 22:06

1 Answer 1

47

I'm not sure if I would attribute much to that article. Bullfrag is a well-known clickbait site which is all about cross-linking and advertisement, not really about the content. Likewise the number of "one million per month in the US alone" seems rather dubious.


The Numbers Presented

So let's dig a bit into MS sales numbers around that time:

  • Wikipedia records sales in its Windows 3.0 article as
    • total sales of 10 Million copies during its 25 months of existence and
    • 2 Million within the first 6 months.

The numbers are credited to Jamie Lendino's 2015 article on ExtremeTech for the 10M and a citation of Edstrom and Eller's book Barbarians led by Bill Gates for the 2M.

The same 10M of total Win 3.0 sales can be found on MS' own site in Michael Desmond's Editor's Note - The Road from 3.0, originally published as part of MSDN Magazine's 2015 Special Windows 10 issue.

Now, that 1 million per month sales number mentioned does in fact show up twice, just not for Windows 3.0, but as a sales figure for Win 3.1. When looking at it listed on MS' own 1992 history page, the April 1992 entry mentions 1 Million preorders:

Microsoft ships Windows 3.1 with more than 1,000 enhancements. The new version creates unprecedented user demand, with over more than 1 million advance orders placed worldwide.

While reporting in August 1992 1 million per month sales for April to June:

Unit shipments of the Microsoft Windows operating system have exceeded one million per month in each of the last four months.

Another interesting data point can be gathered from the April 1990 entry of MS' history - the month Win 3.0 got introduced. It records 25 million MS-DOS users within the US:

Technical support will be offered to the more than 25 million users of Microsoft® MS-DOS® in the U.S. The support applies to all OEM versions of MS-DOS. Because Microsoft licenses the MS-DOS system only to OEMs rather than end users, product support has previously been the responsibility of the individual OEM. Now Microsoft will provide supplementary support.

This also indicates that the 10M number for Win 3.0 sales was most likely worldwide - otherwise, 40% of all US DOS uses would have added Win 3.0, which sounds quite unlikely. After all, the huge majority were production systems with existing software and little need to change the OS. It simply doesn't add up, even taking into account that the PC market was rapidly expanding at the time.


Yes, slow adaption in Japan was about character handling, but not what you think.

While 500k (that's only 5% of worldwide sales) does not seem much, it's a lot considering that Japan was, at that time, extremely different, as NEC's PC-98 series ruled the professional market. While being x86 based, their hardware was not PC-compatible and DOS was only an option. Their software grew over more than 10 years, focusing on the specific needs of Japanese language and communication. More so, NEC fought many long court battles against any competition building compatible machines or software.

At that point, it's important to understand that this is not just about encoding, but the way data is entered. Being based on logographic and syllabic (kanji/kana) characters, it's next to impossible having a key for every character. Thus input behaves somewhat like on today's mobile phones. A user types one or more keys and the system responds with an incrementally refined list of characters for the word or syllable needed. NEC did develop their own method which became a quasi-standard. Except, they did not licence it, but fought every competition very hard. Other manufacturers had to develop their own input systems.

That's the landscape Windows 3.0 entered. It not only supported DBCS for Japanese but was also, due to its fully graphical nature, able to display all those fancy characters. Might have been a great chance to help fully compatible PCs increase their market share - wouldn't it?

Except it was up to each manufacturer to implement their own input method. Doing so also resulted in different encoding in files - not at least for compatibility with their previous systems - and in turn, killing all the advantages a multi-vendor OS could bring.

Now while NEC did adapt and introduce Windows 3.0 for the PC-98 in January 1991, it used of course NEC's IME and encoding.

Fully IBM-compatible PCs thus stayed a niche product. This only changed as IBM introduced DOS/V in November 1990 for their PS/2 systems. DOS/V included its own input method, one free of claims by NEC or other manufacturers thus allowing cross-manufacturer compatibility of data files. DOS/V became the killer application for IBM-compatible PC. It was so important that IBM-compatible PCs are to this day known as DOS/V-PC.

DOS/V was based on IBM DOS 4.01 (DOS 5 followed a year later). The input method was designed as an open, licence-free system, so MS, as well as DR incorporated it into their next DOS versions. More important, MS also used it as a base for their MS-IME for Windows 3.1.

Except, the introduction of Win 3.1 for the Japanese market got delayed and delayed. Originally planned to be in sync with the worldwide introduction in May 1992, it only became available a good year later in May 1993. Which is about when the total dominance of NEC started to wane.


3.0 Was a Good Sale, But 3.1 Was the All-Round Changing Point

While 3.0 did bring a usable GUI to the PC, it didn't offer many features besides. In turn, it was not only slow but also slowed down DOS applications which were still a majority. And it was buggy as hell. That's why the generally forgotten 3.0a shipped in late 1990. It's said it contained more than 1000 bug fixes.

Win 3.1 in turn changed all of that:

  • Way more stable
  • Faster execution for Windows and DOS programs
  • Cut&Paste into clipboard using CTRL-C/X/V
  • Drag&Drop
  • The registry
  • Arbitrary screen resolutions
  • Up to 32-bit colour support
  • Multimedia support
  • Native network support

and most of all

  • The Win32s subsystem for 32-bit programs,

Finally, a standardized way to break out of the real mode limitations for program and data. This made Windows suddenly a serious target to port Unix or mainframe applications down to the PC without all the hassles of DOS extenders.


Bottom line:

  • Total worldwide sales of Win 3.0 were about 10 million
  • Thus monthly sales were on average below 400,000
  • 1 million a month sales were only reached later by Win 3.1
  • 500k sales in Japan must be considered a success as closed as the market was
  • Win 3.0 for the PC-98 was introduced as late as January 1991.
  • Only Win 3.1 introduced a single IME for all OEM versions, breaking NEC's dominance

Some postings on some sites are not worth the magnetic domains used to store them.


*1 - Most notably it was first published for NEC's PC-98 line, and only after that for generic compatibles (together with MS-DOS 5.0/V). Honi soit qui mal y pense.

8
  • 1
    On the other hand, Windows 3.0 was successful enough that it caused the NT OS/2 project to become the Windows NT project. (Though NT 3.1 came out the year after Windows 3.1, I believe the decision was made while 3.0 was the shipping version). Mar 5 at 14:31
  • 2
    @another-dave Wouldn't know anyone claiming Win 3.0 being a fail. Troubled yes, but not a fail. It's just that it wasn't the success some call it in hindsight either.
    – Raffzahn
    Mar 5 at 16:01
  • 2
    I remember win32s. It was a pile of crap. I don't think it had anything much to do with the adoption of 32 bit.
    – JeremyP
    Mar 5 at 17:15
  • 1
    I agree; it was a half-hearted attempt to enable writing code that worked on NT and Ye Olde Crappe Windowes, but there were so many restrictions that it was not particularly useful. Mar 5 at 17:33
  • 3
    TrueType font support and a solid basic set of TrueType saleable fonts were also introduced with Windows 3.1
    – Brian
    Mar 5 at 17:57

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .