September 21, 1998
PCs need to break from the past
By Bob O'Donnell
In most cultures, respect for history and past achievements is considered an essential
characteristic of a civilized society. In the topsy-turvy world of Silicon Valley and the
computer industry, however, historical "legacies" are often rightly viewed as
anchors that slow down the pace of progress.
Nowhere is this more apparent than in the design of personal computers themselves.
Current PC hardware and software architectures are saddled with so many self-imposed
"historical" limitations that it's a wonder they have come as far as they have.
Between the ISA bus; the 16 IRQ, or interrupt request, limit; the ridiculous 1MB real-mode
memory barrier that still underlies Windows 95 and Windows 98; the
sounds-good-in-theory-but-doesn't-work-in-reality notion of shared DLLs; and the speed and
device connection limitations with IDE and serial and parallel ports, the standard PC
hosts a veritable minefield of unnecessary technical restraints.
Of course, lots of smart designers have worked hard to develop new technology standards
that overcome some of these limitations, including the Universal Serial Bus (USB), IEEE
1394 (commonly known as FireWire), and the flat memory models of Windows NT, OS/2, Linux,
and other modern operating systems. In addition, the PC 98 and PC 99 hardware platform
design specifications developed by industry leaders Microsoft and Intel (along with input
from many other companies) are an attempt to force the "retirement" of certain
technologies -- particularly ISA. (See my "PC 98
and PC 99 specs offer hope for the future" column for more.)
Nevertheless, in the case of PC hardware developments, in particular, these new
technologies have been slow to take hold. It seems that some people just can't let go of
technologies that should have gone away a long time ago. Although some of the fault for
this delay lies with PC buyers who keep buying equipment that uses outdated technologies,
I believe the primary reason for this situation is that computer hardware manufacturers
have been reluctant to push the envelope. Whether it is fear of upsetting customers by not
providing full legacy support, lack of vision, or something else, PC vendors simply have
not developed the kind of machines that the industry needs to move forward.
Ironically, I think Apple's new iMac is actually the most forward-looking computer that
has been introduced in quite some time. Apple consciously chose to break with its own set
of legacy technologies (particularly Mac serial ports), as well as the previously
ubiquitous floppy drive, and embraced 10/100Mbps Ethernet and the forward-looking USB
standard for its slick new machine. The company took a lot of heat as a result of some of
these decisions, but they unquestionably move the Mac platform forward, and, in the end, I
think they will pay off handsomely.
Over in the PC world, what is even more confounding is that companies that are involved
in pushing PC design forward have also been guilty of holding back progress. Intel, in
particular, seems to send mixed signals in this regard. At last week's Intel Developer
Forum (see "Intel
readies Katmai onslaught for 1999" for more), for example, CEO Craig Barrett
urged PC designers to break with the past and move away from legacy technologies such as
ISA in order to enable faster, easier-to-use PCs. On the other hand, the company announced
several months ago that it had decided to remove IEEE 1394 support from its
next-generation chip sets because there wasn't enough demand for it yet. Since Intel is
the market leader in chip sets, its decision not to include built-in support for this
important new standard will effectively slow down 1394 acceptance. It seems to me that if
Intel really wanted to push PC designs forward and build the market, it would have
included 1394 support anyway.
Similarly, the company has apparently been dragging its feet in efforts to update the
PCI specification that lies at the heart of today's PC desktops and servers. In an
embarrassing blow to the chip industry leader (and developer of the original PCI spec),
several computer vendors -- including Compaq Computer, Hewlett-Packard, and IBM --
recently announced that they were independently working on a 64-bit, 66-MHz version of the
PCI bus because they needed the performance improvements in their servers. (See "IBM, HP,
Compaq prepare updated PCI spec" for more.) Again, from my perspective, Intel
should have been leading this effort and promising to deliver support for it in its
next-generation chip sets.
Taking the steps necessary to break from the limitations of the PC's past is not going
to be an easy task. There is no question that because of the huge installed base of
peripherals and software that depend on that legacy support to function, some difficult
transition issues are going to arise. But if the industry really wants to move to a new
level of performance and ease of use, it is going to have to make some tough, unpopular
decisions. I, for one, am anxious to see who has the guts to do it first.
©
Copyright 1998, by InfoWorld Publishing Corp., a
subsidiary of IDG Communications, Inc. Reprinted from InfoWorld,
155 Bovet Road, San Mateo, CA 94402. Further reproduction is prohibited.
