windows product sku anti-patterns

In Computerworld, Gregg Keizer takes stock of the Windows 10 testing and release rollout regime on the occasion of the 1809 data loss bug.

The jist of the current situation is that Windows 10 Home users, as non-voluntary first-line participants in the Windows update rollout program, essentially act as beta testers for new Windows releases.  Updates are always rolled out to these users before Windows 10 Pro/Enterprise users and indeed updates are not rolled out to these latter users until sufficient feedback from the Windows 10 Home users is in.

As I reflect on how we got to this situation, in hindsight it doesn’t seem all that surprising, as the signposts pointing Microsoft at this approach were everywhere in the computing ecosystem:

  • Microsoft has always been accused by the peanut gallery of essentially beta testing its products on users, chiefly because of a perceived lack of quality of its software products’ early versions, despite having had independent testing in-house.
  • Microsoft has long provided bundled Windows, mainly Windows Home, to computer manufacturers for pennies on the dollar compared to its retail and other direct pricing, in exchange for something of value: the computer manufacturers took on technical support responsibility for those end users for a period of time.
  • Its difficulty delivering working online store infrastructure for Windows Anytime Upgrade and TechNet, as well as brokenness of Windows Help as it shrinks the infrastructure for that and redirects it toward user forums where assistance can be provided by really terrible quality “community experts”, suggest that it is always willing to do cost cutting and suffer through the quality impact.
  • The effective alternatives to Microsoft in the OS world are basically Linux, where it’s not unusual for a distro to break something in a release upgrade,  and Mac OS X, where they just plain remove drivers for older hardware from newer releases of the OS, preventing users from upgrading (and perhaps forcing them to upgrade their hardware).



a scanner darkly

Canon LiDE 20 scanner

It truly does. It was great in many ways: Competitively priced, at about $100, it was part of a wave of competitive pricing that opened up scanning to the average computer owner. Nevertheless, it was faster and higher resolution than even the $500 scanners of a couple of years earlier.  Industrial design was decent, not the usual beige-y grey placeholder of equipment past.  It was relatively light and powered by its single USB connection, making it reasonable to take along with a laptop to where the documents were.

And yet in many ways it was just another such product.  The nature of the computer peripherals ecosystem hadn’t changed — Windows support was a must, its maker would need to provide drivers, and the lifetime of the Windows driver standards for those initial drivers would end soon enough. It practically disappeared in a sea of mostly similar but slightly differentiated models, as was the style at the time.

For its maker it perhaps represented a new tinier gross margin for that kind of product. The time to spend on maintaining drivers for new Windows releases could be spent instead working on drivers for new products that would have new paying customers.  Technically speaking, its maker continued to supply its drivers all the way to Windows 7. But never a 64-bit version.