Gordon Letwin OS/2 Usenet post (1995)

Original Article ↗ Hacker News Discussion ↗

Once touted as a technically superior successor to DOS and early Windows, IBM and Microsoft’s joint OS/2 project is now often cited as a case study in how strong engineering can be undone by misaligned strategy, poor marketing, and licensing politics. Commenters revisit Gordon Letwin’s 1995 Usenet retrospective to trace how the early arrival of the 386 CPU, IBM’s insistence on 286 compatibility and high-end hardware, and Microsoft’s aggressive push for Windows and NT shifted the software ecosystem away from OS/2. The thread also explores the broader legacy of that era—from the origins of Windows NT in DEC’s VMS to enduring resentment over Microsoft’s business tactics and how standardization and developer tooling shaped which platforms survived.

OS/2 vs Windows: technology and developer experience

  • Many recall OS/2 2.x as technically superior to Windows 3.0: true preemptive multitasking, better DOS support, memory protection, and solid APIs for threads, drivers, and comms.
  • OS/2 1.x was constrained by 80286 protected mode and real‑mode DOS sessions; 2.x was largely a 32‑bit rewrite targeting the 386.
  • Some highlight serious architectural issues: not truly multi‑user and a single input queue for the GUI, seen as a dead end without major rewrites.
  • Memory demands were high: 2.x “worked” with 4 MB but really wanted 8 MB at a time when most PCs had 2–4 MB.
  • Developer tools and ecosystem lagged: Windows had Visual Basic, Delphi, Turbo C++ and later strong Win32 tooling; OS/2 had some Borland support and IBM’s VisualAge, but not at the same breadth.

Business, pricing, and ecosystem

  • Lack of applications and weak IBM marketing/OEM deals are recurring themes.
  • OS/2 was expensive (both OS and suitable hardware), while Microsoft kept DOS/Windows/Office licensing cheaper and friendlier to OEMs.
  • Standardization around Windows drew developers and users; OS/2’s ability to run Windows apps ironically reduced incentive to target OS/2 natively.

Why OS/2 failed: competing explanations

  • One view: OS/2 2.x failed because it could run Windows apps, so developers stayed on Windows.
  • Counter‑view: the real failure was OS/2 1.x on 286; Windows 3.x on 386 exploited its weaknesses (esp. DOS multitasking), gained dominance, and by the time 2.x arrived the market was already decided.
  • Several commenters stress IBM’s corporate inertia and misreading of the PC/home market as at least as important as anything Microsoft did.

Microsoft, IBM, and CP/M/DOS history

  • Long subthread revisits IBM’s original PC OS deal: Digital Research’s CP/M‑86 negotiations, QDOS purchase, and MS‑DOS licensing.
  • Opinions differ on whether Microsoft behaved malignly or just opportunistically; some argue Gary Kildall “blew it,” others emphasize how the “went flying” story harmed his reputation.
  • Broader pattern raised: Microsoft repeatedly leveraging partnerships and legal pressure to its advantage; contrasted with IBM’s own problematic history.

Windows NT origins and VMS connection

  • Discussion covers hiring of the VMS team lead, reusing design ideas from a cancelled DEC project, and how this shaped NT.
  • DEC later sued over similarities to VMS internals and obtained favorable settlements and partnerships.
  • NT is seen as qualitatively better engineered than MS’s OS/2 work, and ultimately the line that displaced both OS/2 and Windows 9x.

Miscellaneous reflections

  • Standardization as a double‑edged sword (praised in OS/2 email; compared to EV charging standards).
  • Nostalgia for OS/2 Warp, RS‑232, Win32s, and early USB; plus side debates on typography (two spaces after periods, long “s”).