Off the Beat: Bruce Byfield's Blog

Stephen Smoogen has withdrawn his proposal that Fedora release only a 64 bit version, apologizing and claiming that it "was meant to be absurd." Still, the change is only a matter of time in all distributions. The only surprising thing is that the transition from 32 to 64 bit computing has taken so long.

To appreciate just how long the transition has been, consider this: the transition from 16 to 32 bit in the early 1990s was a matter of two or three years at the most. In fact, the transition was something of a fiat: Microsoft and the hardware manufacturers announced it was going to happen, and it did. Complaints were few, because, back then, only free software supporters questioned the decisions of software and hardware manufacturers.

To the contrary, users were excited. 32 bit computing would be faster and more efficient, everyone said. The transition would be the next big advance.

It was only after the fact that any number of people realized that 32 bit computing was nothing magical, and that programmers had to learn to write 32 bit programs before any benefit was possible.

By contrast, 64 bit computers became the norm on the desktop by 2003, having first appeared a couple of years previously. Yet, twelve years later, 32 bit software remains widespread, although declining. It has only been in the last year or so that some major distributions made 64 bit downloads the default. In fact, even now, a few like Linux Mint feature both 32 and 64 bit images equally on their download pages, although that may be a matter of layout as much as anything else.

Half a transition
So what's the delay? High end commercial operating systems like Solaris were ready with 64 bit computing in the late 1990s, although I suspect that desktop release might have been delayed until users had got over their fear of the millennium bug (Remember that?). Still, once the transition began, who would imagine the the transition would still be incomplete twelve years later? That's several centuries, at least, in computer time.

So far as I'm concerned, the delay has never been satisfactorily explained. About the best I can manage is that, unlike at the time of the transition to 32 bit computers, manufacturers are no longer confident of their ability to drag consumers along with their executive decisions. Consequently, they have continued to support 32 bit software while building 64 bit machines.

Perhaps, too, the disappointment when everyone realized that a 32 bit computer was still just a computer encouraged manufacturers to under-sell the advantages of 64 bit ones. Another disappointment might too easily have turned into a backlash against their brands.
Granted, for the first few years of 64 bit computers, gaps in the supporting software existed. If you wanted to install some software, you had to create a 32 bit chroot jail, and many people couldn't be bothered. Instead, they were content to wait until a few years until the software had caught up with the hardware.

However, I suspect that the real reason was even simpler: 32 bit computing is good enough for most users. Yes, a 64 bit computing is noticeably faster, but as CPU speeds and amounts of RAM crept upward and multiple cores became standard, 32 bit computing became more efficient, too. Just as many people are content enough with slower, inexpensive laptops, so they learned to live with 32 bit computing. It might not be top of the line, but it was good enough, and so the demand for 64 bit computing never developed very far. Nor did manufacturers care much about the partial transition -- after all, they were still selling computers.

Still, interest at last has tilted in favor of 64 bit computing. 32 bit releases are likely to continue for at least another couple of years -- maybe as many as five -- but, at some point in the near future, proposals like Smoogen's are going to be accepted in most distros. Anything that reduces the work load is going to receive serious consideration.

First World Thinking
Larry Cafiero and others argue that such a move will penalize both impoverished people and countries, but I don't see that, myself. A few old Celerons are probably still chugging away somewhere in the world, but after more than a decade, the majority of used computers are going to be 64 bit. It's only the software that is likely to be the issue, and, many distros, such as Debian, are likely to continue with 32 bit support as long as there are developers interested in maintaining it.

True, 32 bit support will no longer be the main priority. However, if you run the Debian version of Linux Mint, or Kubuntu instead of Ubuntu, you learn to accept the inconvenience of occasionally wait a few weeks after the main release for the version you need. It's an inconvenience, not an unmovable obstacle.

The day of main releases being 64 bit hasn't arrived yet. Still, at geological time rates, it is sliding nearer. By the time time it finally arrives, few of us will notice the event.