Compatibility Does Not Matter

With modern virtual machine and virtualization software, why should compatibility matter?

For instance, I've had far less problems running LinuxOs under VmWare than running it natively on a dual-boot machine. And that's excluding the additional problems of exchanging data between the two. The only problems seemed to be a slight performance decrease, and DHCP.

And then there's XenVmm?. Sounds like a dream come true. So again, why should compatibility matter?

---

Compatibility does not matter to some, however, some is not everybody

[There is a long, long, long, long list of companies that went out of business precisely because they decided that compatibility was not important; apparently the lessons of the PC clone wars of the 1980s have been forgotten. Conversely, the IBM 360, (TechnologyStasis comments?) for instance, is still with us today for a reason. The Pentium derived, one backward-compatible step at a time, from the 8086. There are numerous software examples as well.]

It's funny how Microsoft never went out of business despite Windows not being backward-compatible.

• Not entirely accurate. I found my old software worked in the upgrade about 80% to 90% of the time. However, back then people had to updrade their machines so often anyhow that they generally perchased new OS's only when they got a new machine.

What MS provides is a soft-upgrade path, which is a very different thing from backward compatibility. And MS' upgrade path isn't that soft.

Soft upgrade includes:

• switching and updating the bootloader:
  • without bootdisks
  • in a perfectly safe way
• using an old filesystem
• converting to a different filesystem without:
  • backup/restore
  • repartitioning
  • losing any old data
• using old custom applications
• moving old data to the new system

Soft upgrade does NOT include:

• running old applications natively

Microsoft knows this, which is why applications which ran fine in Win98 won't even install under WindowsXP, no matter how much it touts its compatibility layer.

The only thing which provides most of a soft-upgrade path is a virtual machine such as VmWare. In fact, if you care about compatibility then VMWare should be the ONE and ONLY application to target. Conversely, to take advantage of VMWare, you should make sure that you can easily switch all of the hardware the OS is running on.

So you can take your sanctimonious crap and shove it.

It wasn't sanctimonious, it was factual. They teach case histories on these things in business school, because it can matter to business a lot. I also didn't say that it always matters, just that when an incorrect judgement on the topic is made, it can (and has) put a company out of business.

As for Microsoft, you are factually incorrect. To a consumer it might appear that backward compatibility has never been an issue for them, nonetheless in fact it has been a primary concern of theirs ever since they were founded, and has influenced every OS release they've ever done. They wanted rather badly to switch to NT, for instance, as soon as they got NT running, but were forced instead to support two different OS codebases, and gradually over the years, merged code from NT into Windows and vice versa, growing them closer and closer, until now they are finally variants of a single code base (Windows 2000 was the first NT-based version of consumer Windows).

All of that trouble and cost was purely for reasons of compatibility.

Are there literally thousands of examples of lack of backward compatibility related to Windows? Sure! But that's despite their spending literally billions on maintaining backward compatibility where they decide it is important to do so. Other times, for business reasons they introduce incompatibility on purpose (like the infamous always-changing Word document formats) -- and even then, I wouldn't say it "doesn't matter", I'd say they're using the incompatibility as a weapon.

Compatibility is primarily a business issue. If it were purely a technical issue, without concern for what happened to users when incompatibility were introduced, then incompatible changes to OSes and other products would happen vastly more often than they actually do.

So compatibility does not matter technically but does matter business-wise -- an overstatement, but closer to true.

Now, are you sure you wanted to argue about business matters? I had not thought that such interested you so much.

Conversely, if you just mean to say that compatibility does not matter technically, as long as there's some kind of upgrade path, this is much more defensible.

I meant that compatibility doesn't matter to users. I'm a pretty typical user in this way. I don't care about compatibility at all so long as there's a soft upgrade path for my data. And in those rare cases where I do care about compatibility, I want TOTAL compatibility, which can only be achieved using emulation.

The only reason that compatibility of Windows seems to matter at all is because VmWare has been insufficiently pirated. And that it doesn't work on a pre-existing OS installation, though that could be fixed if OS installs weren't so brittle.

---

Operating systems is probably the domain where compatibility is supposed to matter the most. In reality, it doesn't matter at all. This can be determined by a simple inspection of KillerApplications. What do people do with their computers that they actually bought their computer to do, that they would actually buy a computer to do?

Well, it certainly isn't games and other applications requiring a high degree of compatibility. Rather, what people use their systems for are the five killer applications:

• email
  • word processing
• web
• wiki
• P2P
  • music
  • movies
• organizing all of the above

Not to mention the big one for the end-user (and one that is probably the most effective lockout): games. Oh, and the occasional spreadsheet, photo editor, etc. Business users need database engines, application servers, admin tools, and gobs of other infrastructure.

• Business users can go screw themselves. And what part of "it certainly isn't [for] games [that users actually bought their computer]" did you miss? People who care that much about games buy consoles, not PCs.

• That isn't true, computer games and console games allow different things. People who wouldn't buy computers for games may still let them dictate their choice of OS. However that means Windows, period, and attempts at compatibility won't be important. People who do buy computers for games may want something better on their work system, which goes back to the five above.

So if you've got all these things covered, and beaten, then your OS stands a decent chance of succeeding. Until you account for the other network effects of user and developer mindshare.

Of course, if someone's OS does all of those, then why should users switch? You've hit the nail on the head, Richard--users care about applications. They certainly don't care one wit about exokernels, capabilities, or anything else like that.Microsoft, I suppose, is doing a world of good in that they are educating the world on the need for security, through their own products' lack thereof. But many of the OperatingSystemsDesignPrinciples listed have very little to do with the above applications--other than perhaps making developers' lives easier.

• Actually, ALL of them do. You think a word processor shouldn't be elegant, predictable and reversible? You think email and web shouldn't be secure? You think you shouldn't be able to uniformly change the security on an object to determine whether it's a public web page, a private email or a wide open wiki page? Think again.

• Further, users will switch because ALL of the killer apps can be redone until users don't recognize any of them anymore. And so far there is no app that does "organizing all of the above" and lemme tell you, it's a killer application whose time has come.

BeOs was a damn cool OS; and it certainly could do all or the above--and run rings around both MicrosoftWindows and LinuxOs while doing it. Plus, it had one of the coolest development environments since NextStep. And it could be had for free.

And it still is now a dead system. Pity, too.

It couldn't organize anything because NOTHING out there organizes anything.