When we talk about system or application performance, user perception of that performance tends to be the unseen elephant in the room.
User perception is, of course, driven by three things: expectations, software, and hardware. Of these the hardware is the easiest to understand and expectations the hardest, in large part because we often don't admit the real basis for our expectations even to ourselves.
One example of the kind of dishonesty most of us practice with respect to expectations is that we choose tools to suit ourselves while pretending to do so on behalf of our users. In this context Microsoft is king: they aim everything they do at their customer: not the guy who uses computers, but the guy who sells and preaches PCs, PC services, and PC support.
Traff-o-data's first application was an end user application, but that changed even before the name became Microsoft, with the focus shifting to providing the tools for financial success for people acting as intermediaries between suppliers like Microsoft and end users like the New Mexico Department of Highways.
That business focus worked well enough in the early days to become cast in stone as Microsoft's key operational value with the ethical compromises that went into the run-up to Windows 3.0 - not the lawsuits and more or less overt copying, but something far more fundamental: the abandonment of any attempt to honestly serve the end user in favor of providing Microsoft's real customer, the profit or status seeking PC evangelist, with the tools needed to deceive first himself and then his customer.
At the time, the mid to late 80s, lots of other companies, including Apple, Atari, DEC, IBM, and Sun, had working GUIs, but all of those were built on Motorola CPUs with 24bit memory addressing and couldn't run on the 16bit PC.
The x86 products that did exist, notably those from Digital Research and Microsoft, ran best on faster 80186 machines like those from Tandy and in the larger memory spaces made available with PC-DOS emulators under SunOS, AT&T Unix, and SCO/Xenix. Everyone knew the i80386 was coming, and would be significantly faster despite Intel's odd decision to continue the 16bit memory block - but Microsoft needed something that would make the installed base look good, and most of that consisted of pre 12Mhz 80286 gear with 640K of ram or less.
The solution they came up with was literally brilliant: Windows 3.0 shipped with one very highly optimized function: to first paint the frame for a new window using a bright primary color, line that with a contrasting color box, and then fill in the interior in preparation for the text box requested by the calling function.
For most people, most of the time, peception of frame drawing time dominated all perception of total time - meaning that people simply missed most of the delay in getting the contents of a new window up on the display, and therefore perceived Windows 3.0 to be much faster on their hardware than it actually was.
As a strategy this was so successfull that it became the basis for a visual exploit: the productivity crippling Windows mannerism of splitting up real work among lots of single step pop-ups all of which interupt work focus and flows -and indeed Aston Tate later built it's (non relational) relational database into a world-wide development tools empire on little more than the ability to extend Microsoft's fast frame technology to differentially color multiple rows inside a frame.
Setting aside ethical issues on using a perceptional characteristic inherited from our days as predator dodging monkeys against the customer's ability to judge actual response time, you'll no doubt have recognized this both as the origin of Microsoft's characteristic colors and of the Potemkin boot strategy under which XP completes booting well after issuing its initial login prompt. What you might not have recognized, however, is that Microsoft's double whammy benefit came from the self-validating effects of the misperception, not the misperception itself.
People look for validation - and the higher the emotional stakes the further we'll go to get it. My divorce lawyer is smarter than yours, and the sicker I am, the smarter my doctor is relative to yours - or, as Festinger documented in his earliest work on cognitive dissonance, people who bought Fords tried to validate that decision afterwards by actively seeking out Ford ads while ignoring Chevy ads.
A more Nixonian phrasing, however, would be to say that if you have the customer by his dollars, his heart and mind will soon follow - and that's exactly what happened with Windows 3.0: the perceived speed of response validated financially motivated seller decisions to get into the x86 community and correspondingly reduced the barriers to believing that the Wintel PC was somehow competitive with workstations from Sun or PCs from Apple.
What evolved from that is what we have now: a situation in which corporate PC people have their careers bound to the infrastructure in place, exploit Wintel's dependence on endless rounds of desktop hardware and OS upgrades to justify their own jobs, and finally use desktop lockdown to achieve end-user lockout -denying them any role in choosing or operating "their" desktop computing tools.
That's the evolved product - more complex, but not functionally different, from the basic recipe for Windows 3.0's marketing success: take trailing edge products, focus the pitch on a financially motivated decision maker who desperately wants to believe, and ignore the interests of the actual end-user as irrelevant to the sales process.
This entire strategy depends on taking the end user out of the decision making process - because you have to care deeply about computers to be affected: and that means careerists, hobbyists and PC "journalists", but not the average business user.
So what do real users, the clerks and others who just want to do their jobs, really want from us? As I said earlier, that's the invisible elephant; the one everyone pretends isn't in the room - and this week's topic.