OS/2 e-Zine! -

OS/21st - by Sam Henwrich

Thoughts On The Future Of The Perfect OS

Summary: Assuming the minimum guaranteed by Moore's law, how might the nature of an operating system change in the next 12 years?

What do you think computers and operating systems might be like in 12 years? In this time period, taking us to the year 2010, lets assume at a minimum that Moore's law keeps on track and produces twice as much computing power for the dollar every 18 months. At the end of 144 months, the power of computers should have doubled 8 times. What do you do with all that power? Here's some thoughts:

Applications That Never Close

Hitachi has already produced the world's first 1 gigabyte memory module, using thirty two 256-megabit chips. This means computers are at the point of having almost as much RAM as hard drive space. That prompts an interesting idea: What if all the applications we run were loaded into RAM at install time and never exited?

Consider that in 12 years the average computer could easily ship with 32 gigabytes of ram as standard, a power users machine might be equipped with 64 or 128 gigabytes of just RAM. At this point the ratio between data size (audio and video files, multimedia files etc.) and application size (the programs and their code libraries) will be far greater than it already is today. It won't be unusual to work with a 50 gigabyte movie with a 50 megabyte application, for example. And twenty 50-megabyte applications would be a drop in the bucket for a machine loaded with 32 gigabytes of RAM.

With all your applications loaded into RAM at the same time, most of them staying dormant for days or months, the wait time to load them will be completely eliminated. Instead of starting an application, you switch to it. There is never any application which you can't switch to. The concept of actually loading an application from the disk whenever you need it will become as outdated as the old DOS days of having to close one program first before you start another.

To be honest, the operating system might divide programs into two or more classes, the basic two being Applications and Utilities. Applications never close, they have no 'quit' or 'exit' option, they simply minimize when you're finished with them. Utilities will have the option of closing, however. It's like the difference between furniture and your vacuum cleaner -- Furniture you keep out all the time, but when you're finished with your vacuum cleaner you always put it back away in its cupboard.

OS/2's Workplace Shell has sort of promoted this idea a little bit by the way it treats the launching of objects. Double clicking on an application or document icon will launch it if it isn't already running -- but switch to it if it is already running. Clearly the standard task-switching bars and window lists we're used to today won't be nearly sufficient enough when everything is running at the same time -- but Warp's Workplace Shell is already the best interface in the industry for organizing thousands of applications and documents. While I'm expecting the WPS will change a lot in 12 years, it's not as if we've got any major hurdles to leap over in that time.

The Elimination Of Applications, And Documents That Never Close

The counter thesis to the above is to eliminate all standalone applications entirely, and instead shift the "never close" concept to documents. What we know of as the relationship between applications and the operating system -- that each are separate entities, will disappear in favor of a new model: That the operating system is your one and only application. Instead of buying a program that lets you edit letters or spreadsheets or video streams, you buy a Document Type that gets merged with the operating system.

The Document Type is simply a CODEC or set of modules that give the operating system the power to load, save, and define how to set up the user interface to edit that document. Ideally the Document Type would be minimal in function and the user would buy extra features separately or in a bundle. These might be a function library for a spreadsheet, spell-check and thesaurus dictionaries for a word processor, filters and special effects for a paint program.

Like with applications, the operating system might have to split documents into two different classes; mundane and special. Special documents would be left on the hard drive and loaded only when needed -- like the current model. Mundane documents are small enough and routine enough to stay in RAM all the time. The OS may also need to do some behavior analysis of the user and archive infrequently used files and Document Types to the hard drive.

The Computer That You Don't Switch Off

If all applications or documents are to stay in RAM all the time, it'd be a pain in the rear for the computer to load them all up every time you switched it on in the morning, wouldn't it? That's why in 12 years you'll see computers that don't switch off, even after you pull the plug out. Like laptops of today, which can switch between battery and wall power instantly without interruption of service, so should the desktop computers of the future.

All mission-critical environments have UPSes (Uninterruptable Power Supplies), but even a small UPS for a personal computer is fairly cheap -- about $80-90 -- so it won't be long before they get merged into the main box just like disk drives and modems did. Plus, in 12 years, battery technology and energy efficiency will have advanced to the point where you could run the machine for days or weeks without needing to recharge (I'm also assuming everyone will be using flat-screen displays by then, too.)

During the times when you're not using the machine, you can either put it into energy saving standby mode or let it time-out on its own, just like a screen saver. We'll come to think of our computers as "Instant On" machines, and the boot-up process might happen only in the factories.

And The OS That Doesn't Crash

So, now we come to the ultimate conclusion: If applications or documents are to stay loaded in memory all the time, and the computer is never switched off or rebooted, we need to have an operating system that does not and cannot crash. What we need is an OS that does not have a 'crash' condition, and you make it by simply not writing that condition into the operating system's code.

This sounds about as ridiculous as the theory that you can make an indestructible airplane by building it out of the same metal that they make black boxes with. I mean the black box is always recovered intact, even in the worst airline disasters, right?

But it's not as far fetched as it sounds, because 99% of all crashes, lockups and unrecoverable malfunctions are due to conditions the programmers didn't anticipate happening. The other 1% are hardware problems. (By the way, if any of you out there think this ratio is off, there's a nice talkback forum linked to at the bottom of this article, hint, hint.) This means that in theory, better programming practices, peer review of source code and reuse of stable code should get us to the point where the operating system never fails. Linux, as an example of a heavily peer-reviewed operating system, is already nearly crash proof. As a result it's very popular among companies wanting to build sealed boxes and consumer devices (price not being the main factor for these companies).

Could it be that OS/2, highly stable but nowhere near crash proof today, will be in 12 years? In 12 years, IBM may have decided to open up the source code after all, or their decision to stabilize development of the kernel will mean 12 years of service packs has given us a true, bulletproof OS. Personally I favor the former, since the latter means nothing exciting to write about for a decade.

When your machine does fail, hopefully only once every couple of years, then you know that the problem must be truly serious and it's time to take it down to the shop.

Benefits and Costs

All of the benefits are obvious. Instantaneous loading (switching) of applications or documents, "Instant On", and reliability you don't even have to think about. But what will be the costs?

The number of applications you can install and documents you can create would be limited by RAM. But then, they're already limited by hard disk. It's possible though that in 12 years, the difference between long term storage (what we now use hard drives for) and short term storage (RAM) will disappear with the development of fast, cheap, nonvolatile RAM. If this day comes, the hard drive will fall back into specialized scientific and engineering use, and the only thing you'll have to worry about upgrading is RAM and possibly the CPU.

If you have your own ideas for OS/2 and what the future of computing might be like, talk about them in our Hypernews forum.

Sam Henwrich is an upstate NY OS/2 user. He can be reached at henwrich@yahoo.com.