3 ms·
Come off it. Making some hardware and writing a device driver is hardly fragmentation. If the SATA drive didn't run on Windows the day it shipped, there would b
by konad 16y ago
Come off it. Making some hardware and writing a device driver is hardly fragmentation. If the SATA drive didn't run on Windows the day it shipped, there would be no point shipping it.
- CWuestefeld 16y agoI think you dismiss my point to quickly. You comment on my point about disk drives. However, the differences between mass storage media is most certainly a factor that authors of many applications need to consider. Capacity and performance characteristics are relevant in video editing, for example. Optical disc burning software must consider the quantity, performance, capacity, erasability, etc., of the drives (and discs) present -- and must deal with different driver interfaces as well. The difference between graphics adapters is an extreme difference. Authors needing advanced graphics capabilities must be able to handle different capabilities (resolution of course, but also shaders, etc.), and in extreme cases, GPU code downloaded to the adapter. They may have to deal with different software interfaces (DirectX vs OpenGL). And the handling of multiple monitors is a common question these days. Even CPUs, as glibly dismissed, have far more differences. For some kinds of applications the developer may need to deal with the presence of SSE2 or SSE3 instructions, for example. There really is a tremendous amount of fragmentation in PC-compatible hardware. And your reply about "writing a device driver" is, of course, part of the answer. When the variance is sufficiently important, developers create abstraction layers so that they don't need to worry about it -- that's the point of things like OpenGL or TWAIN.
- elblanco 16y ago>Making some hardware and writing a device driver is hardly fragmentation. What exactly is your definition of "fragmentation" then?
- konad 16y agoChoices that requires commitment and are hard to roll back.