First of all I can tell you it is a lot harder than you seem to think even
for a skilled user. The part numbers don't match the chipsets. You'd know
this if you have worked with thousands of different parts. We run into this
issue constantly.
But yea- it's not impossible to put together a system that works pretty well
for a skilled user. It just takes a bit of work + a little know how. I would
not disagree with that general assessment. Most skilled users though aren't
going to want to mess around finding the right part. They want it to work the
first time and even when you have all the known good parts things still fail
frequently enough for it to be an annoyance for most people.
There are also other issues with older systems/parts. Now you might be fine
depending on the source of the parts. Old parts that have never been used and
are still sealed might be fine. However the older parts in most cases tend to
have issues. There "refurbished" parts or from systems people threw away
because they had problems with them. The might work some of the time for
instance-but that just wastes more time (of which I generally put a value on
even for people making near minimum wage- in the USA anyway). These
inconsistencies can cause huge amounts of time to be wasted. It's bad enough
with the defect rates of new parts where 90% of people are at least satisfied
with them.
Plus- the costs would drop if more people bought free software friendly
hardware (where they can- this isn't easily the case for people in places
like Africa, Asia, South America).
Honestly it probably won't matter. What is likely to reduce the costs are
large corporate sales. But don't be mistaken that every sale doesn't move
things one step closer.
Just to give people a general idea of why we should care more. Some
projections of sales to the GNU/Linux community for USB wifi adapters:
500,000. If even a small fraction of those sales were to one company that
would be millions of dollars a year. That is money you can do something with.
The Free Software Foundation is running off a mere $300,000. The problem is
people are buying from companies who don't have any stake in it. The hardware
works merely by chance. Not because companies are spending money to make it
work. That might be good enough for tech savvy users although it is still in
ones disadvantage to underfund development. What you get from that is people
working only on 'interesting' projects. You get fewer bug fixes, etc.
