Driver me crazy

Drivers: You can't live with them, you can't live without them. They're the glue that holds hardware designs and operating systems together, a glue that can be as strong as steel or as weak as rubber cement on a hot day. In more than 20 years of network computing, we haven't straightened out the mess associated with these pieces of code that run our computing lives. They're tough to document, tough to install, and the devil to keep fresh.

Once a great hope for stability, driver certification programs have become just another profit center for operating system makers, say hardware makers, and they're partially right. We all suffer for lack of cross-industry standards for everything from driver nomenclature to traceability.

For example, while Microsoft, the Linux Standards Base and others have set engineering standards for revision control numbering, driver makers can do as they please, suiting themselves rather than the needs of network engineering personnel.

To continue reading, register here and become an Insider. You'll get free access to premium content from CIO, Computerworld, CSO, InfoWorld, and Network World. See more Insider content or sign in.

Drivers: You can't live with them, you can't live without them. They're the glue that holds hardware designs and operating systems together, a glue that can be as strong as steel or as weak as rubber cement on a hot day. In more than 20 years of network computing, we haven't straightened out the mess associated with these pieces of code that run our computing lives. They're tough to document, tough to install, and the devil to keep fresh.

Once a great hope for stability, driver certification programs have become just another profit center for operating system makers, say hardware makers, and they're partially right. We all suffer for lack of cross-industry standards for everything from driver nomenclature to traceability.

For example, while Microsoft, the Linux Standards Base and others have set engineering standards for revision control numbering, driver makers can do as they please, suiting themselves rather than the needs of network engineering personnel.

The hope was that this wouldn't happen. Novell started one of the first network hardware certification programs, and Microsoft and others copied it. In the land of Linux, BSD (and Darwin), and Solaris, things are only slightly less chaotic.

What's the root of the problem? Driver certification is expensive for hardware vendors and it adds significantly to time to market cycles. Take six current versions of Windows, two of MacOS, three of NetWare, two kernels of Linux and so on, and you can understand why hardware vendors reel at the costs of supporting their products under the aegis of various driver certification programs.

One of the great pains of deploying a new operating system is that these new operating systems don't always come with the latest and greatest drivers in the distribution CDs. Having fresh drivers would give you a pleasant out-of-box positive experience - or at least that's the hope.

Instead, the onus is on the buyer to chase down driver software for each device in each system. The purchaser/installer/administrator then must hope the driver is correct for the platform, and with some luck, might have the distinction of being certified by the operating system vendor and authenticated with a digital certificate.

In our recent lab tests, whether on Windows, Linux, MacOS or xBSD platforms, the biggest problem has come from driver updates and synchronization. What could have been a miracle in stability, reliability, and even cost-effectiveness has been dashed against the rocks of complacency.

Lacking standards and protocols for device drivers and downloaded files, we'll continue to have entropy. If you reject unauthenticated drivers or those not the result of an operating system vendor certification/authentication program, you'll often be left high and dry.

What are your options? Complain loudly. Buy elsewhere. Or demand that everyone work together to get a fair and reasonable methodology for fixing the problem. Lacking that, continue to buy aspirin in bulk.

Henderson is principal researcher for ExtremeLabs, of Indianapolis. He can be reached at thenderson@extremelabs.com.

Read more about software in Network World's Software section.