Software Version Numbering After 2000?

apsmith wrote in wondering what software makers (like Microsoft) will be doing with their software versioning now that we've passed the year 2000 milestone. It's a humorous look at software versioning and it poses some interesting questions. What do you do when you cease using a sensible versioning system in favor of marketing hype (ala "Windows 2000")? Click below for the full text.

apsmith asks: "As I just heard that Microsoft is naming the next version of its database SQL Server 2000 it got me wondering - what happens to all these software products with big "version numbers" in a couple of years when 2000 seems like ancient history? Will we see more factor-of-20 leaps to Office 65535, Windows 1048575, etc? Merely modifying the fourth digit of the version number seems too insignificant to make upgrading seem worth the hassle - does Windows 2008 catch your eye any differently than Windows 2005?

It's not just Microsoft products that seem to have written themselves into a corner with high version numbers, though they are probably the worst. But even Emacs is up to version 20. Sun pushed Solaris from 2.6 to 7. RedHat at 6.1 is somehow way beyond the Linux kernel. At the other extreme is the model that Donald Knuth took for TeX, with the version numbers slowly approaching Pi (the latest teTeX distribution has TeX version 3.14159) but TeX hasn't changed much in the last 10 years either, so a lot of extra pieces have evolved around it to keep it functional.

In the real non-hyped world it seems any version number over 5 or 6 implies it's about time to switch to a new product or start over from scratch. There are countless examples - from recent history think of libc6 -> glibc2 (a bit of a mess there), HTML 5 -> XHTML, or perhaps even Netscape 5 -> Mozilla. Or is that just a geek's view of the universe? How should we be numbering our products these days? And what is Microsoft going to do after 2000? "

Comments

Switch to hexadecimal

[Imperator]

If your product has a single digit version and you're running out of integers, switch to hexadecimal. You can say you were using hex all along. Or do what Apple is doing, and move to Roman numerals.

Re:2002+

[Ranger Rick]

Windows 2001 is obvious and an accepted futuristic number.

Yeah, I can see it now:

"Open the CD-ROM bay door, BIL."
"I'm sorry, I can't do that Dave."

Year Versioning Makes Sense

[VAXman]

I think year versioning makes sense, especially since software revisions tend to be every year or every couple of years. It is easy to pinpoint exactly which products go together, and how up-to-date your package is (which is good from a marketing standpoint). On something like VMS, the current OS version is 7.2, the current DEC C is 5.6, the LSE editor is something like 11.2. The version numbers are all out of sync. A number like 7.2 is meaningless. With Linux it is even worse, especially with the kernel version and distribution version out of sync, and all other product versions out of sync. For Microsoft products, it is obvious that Windows 2000 is the most up to date version, that Office 2000 goes with Windows 2000, etc., so I think that's good. I hope that VMS and Unix products go towards year versioning. The bottom line is that a year version means something, while an ordinary version is meaningless and arbitrary.

I also hope hardware goes to year versioning also. What does 21264 mean? What makes Pentium II newer than Pentium Pro? It would be much easier if it was "Pentium 2000", "Alpha 2000", etc.

Cars use that convention. As computers become consumer type appliances, it makes sense that they use the conventions also.

Re:Year Versioning Makes Sense

[Plasmic]

Year Versioning is only half of the solution:

Even Windows 98 has a Second Edition and countless updates, Windows NT/2000 have always had build numbers, and the list goes on and on. What it amounts to is that 'year versioning' is the marketing/public side of versioning, and the real versioning takes place in the alleys, with my Internet Explorer 5's version at 5.00.2919.6307 q246094 (really!), and Office 97 at Service Release 2b and the Jet 3.5 update.

What it amounts to (as far as most people are concerned) is that year versioning is good for when I'm talking to my relatives so that all I have to do is say "does the screen with clouds at the beginning say 95 or 98?" to begin troubleshooting their problems at Christmas get-togethers, but when I'm talking to computer-savvy folks, things like "Slackware 7.0" don't even begin to describe what's really inside my box.

My thoughts...

[elixir]

I like the way many Linux distributions code name releases. (Manhattan, Slink, Potatoe, etc.) Why can't we do that more often? I think it's much easier to remember Cheese vs Potatoe rather than say 5.1 vs 5.0.1. I'm not saying loose the numbers altogether, but just emphisize the code name more.

Either code names or we can use another industries system... (GT, VR-4, SX, GS, etc.)

Spice it up!

Mozilla has it right

[Money__]

Mozilla has it right.
I want nightly builds.
Access to the bug database.
Browse the source.
My advise to software marketeers?

Release early and release often, and above all, be open.
_________________________

Year-based naming

[retep]

Year-based naming may be nice and easy to follow. But you have to remember that when you say Windows 95 it "feels" old. Tieing the name to the year the software was released is just a way of making sure people remember how old their software is.

Experiences from a *REAL* computer user.

[slashdot-terminal]

I have had various problems with some packages in debian unstable. Because of ongoing levels of development substantial changes can occur from say
package-1.0.4-45 to package-1.0.4-46 with various security fixes and improvements etc. Upgrading to the newest version will often times allow you to use the latest features that the community around you uses. If you don't upgrade I think that some people are just afraid or clueless. Just like some idiots who still run dos version say 6.0 instead of 6.22 despite various changes ( I have seen them). Running say kernel 2.0.33 differences from 2.0.34 may not be in the actual changes to the kernel but from various contemporary changes in patches and add on features that the community will add to 2.0.34 and not 2.0.33 because it is the latest thing.

Year versioning sux!

[Roadmaster]

Back when word leaked out that the then next release of Windows would be named after the year (95), there was a lot of talk that Microsoft would have to back off and return to standard versioning schemes. Those who said so cited the apparent failures of other attempts at year versioning (Mace Utilities 90 which later faded out of sight, and Illustrator 89, which was followed by Illustrator 5.0).

5 years later, it seems as if Microsoft succeeded in doing things their way, and now everybody is wondering whether this is where the industry will be headed.

I for one think year versioning is stupid. It doesn't, as Microsoft and others claimed, help customers unambiguously identify a product's latest release. Take a look at the (at least) four different versions of Win95. A major.minor versioning scheme would have been better for identifying the latest release.

Then you have the "year" releases of other products, and then you see the "clearer" year versioning scheme fail as you see people talking about "windows 97" (since a big "97" pops up when they run Word or Excel from Office97) or Windows 2000 (same thing, except they bought Office 2000). It makes knowing *what* version people have a nightmare.

Also, as with cars, you have year-named stuff being released before the year. How does a common mortal know that office 2000 was *not* released in 2000? How will it help when, in 2001, say they release Windows 3500 and Office 16384? And since they are no longer sticking to the "name it for the year it was released" scheme, how do I know whether my version of Office 2000 is the latest, or has been superseded by "Office 2048", released by microsoft heralding the coming of power-of-2-based versioning schemes?

I say just use the tried-and-true major.minor.revision scheme.. it has worked well for years.

Sun is almost as bad

[THB]

I hate the way that sun versions their product. SunOS 5.7 = Solaris 2.7 = Solaris 7
This becomes worse when you are also dealing with people running different verions of both Solaris and SunOS.

What seems to have happened with both MS and Sun is that the marketing people are now controlling the versions, not the developers.
Solaris 7 sounds (at least in their eyes) more mature than 2.7, and microsoft follows the same logic with with windows 2000 over NT 5.0.

I think that microsoft will stick with the year system at least until the two braches are merged, which was what was supposed to happen with 2000, oh well.

Solaris 2.7/7 name

[ChrisRijk]

Sun have a page & FAQ about this change. It is basically just marketing. I think their statement about how that the '2.' part is basically redundant because they have no changes planned that would justify a '3.' release, is pretty fair. So, instead of Solaris 2.5,2.6,2.7,2.8, they did "Solaris 7", the next is "Solaris 8", to be followed by "Solaris 9".

btw, unlike what some people do, the name only shows up in marketing/documentation/logos. With 'uname', the OS reports itself as being SunOS 5.7! (Solaris 2.X is SunOS 5.X) Backwards compatible with 2 levels of marketing re-branding ^-^

I don't particularly care what Microsoft do... btw, The Register has an amusing article on "Microsoft Year 2000".

Year Versioning DOES have its place

[ShadeTC]

I'm going to get a lot of argument for this but I think in certain situations that year versioning is a "Good Thing (TM)"

Any product that has a built in shelf life or has major changes that are tied to the year of release should have year versioning. Mostly this is for financial software such as TurboTax (TurboTax 98, TurboTax 99, TurboTax 2000) TurboTax is useless for the most part the year after it is released, due to changes in the IRS tax code and the forms, and the calculations etc (You can bet Intuit is NOT a supporter of the flat tax. TurboTax is a little cash cow)

But I agree that for things such as Office and Windows, we should see an actual numbering system.

I like the way that BeOS does it RX.Y.Z Where X is the main version release, Y is the minor upgrades, and Z is little updates like drivers and bug fixes (very similar to other companies use of the "SE" title :)

TC

Parts have versions, systems have names.

[Above]

I think the key part that is being missed is that parts have versions, but systems have names. This is particularly true when the systems have interchangeable parts. I think the car analogy was a good one, so I will go with that.

When you design a part, like a spark plug, you give it a version number. These probably are some take off on the traditional software scheme, with "major" and "minor" revisions. The first three copper ones are 1.1 1.2 1.3, and the first three platnum ones are 2.1 2.2 2.3. This makes a lot of sense, and tracks the evolution of a spark plug nicely.

When you use those parts in a system, there are a wide variety of version numbers, and they don't mean anything relative to each other. Version 2.1 of the spark plug was not designed at the same time as 2.1 of the muffler. So, you name the system (or version it, if you must) as a whole, and leave the individual version numbers as something to be droned on about in the detailed spec.

This works out nicely. I go buy a 2000 Viper (hehe, i wish) and it comes with version 2.3 sparkplugs. Later, when they make a better one I can go to 2.4...or I can swap out for version 1.7 of another manufacturers design, which are better. It's still a 2000 Viper.

Software works too, Red Hat "6.1" (a name, not a version) is made up of parts of all different versions, and that's ok. We also all know you can interchange at least some of those parts, and update it individually.

So, I expect all "parts", eg software components to have monotonically increasing versions numbers like they always have. I also further expect marketing types to come up with cool names for new products that let me know one is better than the next. Cheetah is faster than Baracuda is faster than Wren, you know... but all those disk drives are made up of many versioned parts.

I think the "2000" name is a fad, and will quickly fade now. I expect the next name to appear equally stupid to many of us, but the lemmings will buy it anyway.

My Mom: Yea, our office has the Microsoft 98

[Money__]

What follows is an actual conversation that took place before my dear mum sent me a file, and why year versioning isn't good.

Mom: Yea, our office has the Microsoft 98
Me: Mom, that's the operating system, What word processor?
Mom: Word? yea..we use Word. Word Perfect 98.
Me: Is that Word, or Word Perfect?
Mom: I don't know, it's the one with the little squiglies.
Me: They both use that interface for spell checking, could you just copy and paste the text into the email?
Mom: huh? Why don't they all just use one system.

And in one bright, shining moment in my mums 'puter understing and growth, she wraped her mind around the value of standards, and open documentation, and I wanted to jump for joy.

In summary, Year versioning is a confusing marketing ploy that's not good for developers or end users.
_________________________

Re:I'll give you a clue:

[SEWilco]

On a Unix file system, "/." also means "the directory which is the current directory within the root directory". Or "When at The Beginning, You Are Here."

Version # inflation - it's sorta like pinball.

[seebs]

I dunno how many of you play pinball, but I remember games where a score of a million was a huge result, and I vaguely remember games where a score of 100,000 was impressive. I saw a TV show once where a score of 45 was pretty good.

By contrast, there's periodically speculation about what kind of changes would merit a major number change in NetBSD. 1.0 was released in 1994, just over a year after 0.8. 1.1 was a little over a year later.

The most recent NetBSD release is 1.4.1; -current
is called "1.4P". Somewhere in there, my i386
converted to ELF, got CardBus support, got a complete rework of the concept of console drivers, got a framework for multiplexing input devices (to make USB keyboards and mice relevant), and got soft updates. None of that justifies a bump to 1.5, apparently.

The bump from 1.3 to 1.4 was a *COMPLETELY NEW* VM subsystem, new compiler, and a dozen or so other features.

I guess it's just an island of sanity. :)

Redhat is Satan!

[Inoshiro]

Now that I have your attention :-)

Redhat Mother's Day release +0.1 -> Redhat 4.2

This happened in the same amount of time it took to go from Slackware 2.3 -> 3.x.

So, of course, Redhat had managed to get up to 6.0 by the time Slackware hit 4.0... Jumping 4 whole numbers, while Slackware went up < 1. Then going up two more, while Slackware hit 4. Sigh. Where's my Bob Young voodoo doll? ;-)

---

Re:Microsoft Version Numbers

[Inoshiro]

Win95 = 4.0
Win95 A = 4.0 sp1
Win96 B = 4.0 R2 (FAT32)
Win95 C = 4.0 R2.5 (USB!)
Win98 = 4.1
Win98 SE = 4.1 sp 1
Sigh.. Way too much time working as a tech supporting bad software ;-)
---

On the obscuring of version numbers in Windows

[xdc]

Version numbering used to be simple and sensible, but lately -- in the Windows world, at least -- it has become incomprehensible. It used to be that you'd have the major-point-minor version. Since some minor versions amounted to tweaks and bugfixes, they would increment the version by 0.01.

Now observe what has happened with Microsoft products. Microsoft started using years instead of version numbers in its product names, but did so inconsistently. Instead of Windows 96, it was Windows 95 OSR 2. Instead of Windows 99, it was Windows 98 Second Edition. But it gets worse...

Windows NT 4.0 has undergone some very significant updates that certainly merit at least new minor version numbers. A lot of NT software won't even run without these enhancements and bug fixes. But instead of calling it Windows NT 4.3 or at least 4.06, we have Windows NT 4.0 Service Pack 6.

Internet Explorer versions are the most baffling. Various schemes are mixed and matched, so we end up with things like IE 4.01 SP2, which is slightly different than the other IE 4.01 SP2 that was released before. And the version number embedded in the software is some crazy dotted quad beginning with 4.72! (I probably erred a bit on the exact version and service pack numbers in the preceding example, but I am not making this up!) The latest nonbeta IE is 5.01, but when I check its version number, it says 5.00.2919.6307! Why not 5.01?!

I suspect that this confusion may ultimately be part of a long-term plan of intentional deception. From Microsoft's point of view (as I see it), users should be kept out of system internals. In fact, operations should be so transparent that one will not and should not know whether something is coming from the local hard drive or from an Internet connection to Redmond. Both good and evil updates will be quietly slipstreamed through users' always-on broadband connections.

Already, Microsoft Outlook hides the email addresses of email and newsgroup messages, showing only names like "John Q. Random". Though it is possible to find the return address via a troublesome multistep process, I have found no way to disable this misfeature in any of the option dialogs. Sure, these are just small annoyances. The end result is that users will have even less of a clue how the systems their livelihoods depend on works. Although I'm for intuitive interfaces and ease-of-use, I deplore the dumbing down of people.

I think the future holds some corrections. Namely, the use of codenames or other product names in place of major version numbers. It seems that at this rate, the world may see Microsoft Windows 2003 Professional OEM Service Release 2.0 Service Pack 3. But I think it will probably be more like Microsoft Windows Neptune Server, and if you want to scrounge around in DLLs, you can find out each component's version number.

Using Letters For Version Numbers

[grantdh]

While working as a developer for a company in Australia, we found that users of our Insurance software were comparing versions when they met (the Insurance industry is rather incestuous and many used our software). We wound up getting users stating that "They're on version 9.121 while we're still on 9.049 - we want an upgrade!" despite the fact that they had no idea what had changed between the versions (usually small customisations for new/changing clients, etc).

To get around this, we started to use a main version plus two letters (eg: 12 AS, 12 TC, 12 BS, etc). The letters were not allocated in any order but were different in each version. This let support staff ask clients their version to check for known issues but dramatically reduced the number of "I'm using obsolete software" calls.

Of course, we had chart to show which versions were assigned to the various letters and there was also a command line call to get the full version.

A side effect of all this was that people started to "name" the versions (eg: 12 AS was known as the "Arnold Schwarzenger" version, etc). As a new version was released, we'd go through the two letter combinations still available and figure out names to use. Sad but true...

Bill Gates: "Windows 2001"

[Noke]

I'm watching the Larry King Live interview with Bill Gates and he mentioned the version of Windows following Windows 2000 would be Windows 2001. (9:56Pm EST)

WinNT5

[nordoff]

"...In the real non-hyped world it seems any version number over 5 or 6 implies it's about time to switch to a new product or start over from scratch. There are countless examples..." Windows NT 5 -> Linux 3.0