12 Jun
2013
12 Jun
'13
6:15 p.m.
On Wed, 12 Jun 2013, TeoZ wrote:
I think many developers quit trying to have their apps
run on old average
hardware a long time ago and say so in the specs for it.
A well defined spec clearly identifies what range is suitable. It is then
the developer's responsibility to work right for everything anywhere
within that range.
A tightly defined spec makes for an acceptable range.
If they define their target platform as "ALL PCs" then that is what they
need to work with.
Setting a target platform spec that is too all-encompassing creates an
obligation that they can't meet.
If the software supposedly runs on a ratty old Pentium 4 with 512MB of RAM
running a locked-down Windows XP that won't even let me install PuTTY or
an editor, then that is what the developer needs to deal with. Not all of
the time, but at least enough to have a feel for what the hell is like.
Again, a proper development environment heeds to have the entire spectrum
of the target platform. High end machines for developer productivity AND
POS to do REAL, not just token, testing. MICROS~1 never understood that
seeing if it booted, and running a single game of Solitaire (their only
thoroughly tested product at the time) doe NOT really constitute
"testing". Making the staff USE 286's for their word processing and
other "lesser" tasks would have improved their awareness of the platform.
Microsoft stopped testing 3.0 with 8088 long before they stopped selling
it for that, and stopped testing 3.10 with 286 long before they stopped
selling it for that.
Properly planned testing might significantly reduce the amount of testing
needed. Badly designed testing does nothing but waste everybody's time.
Number of games of Solitaire played is not the right metric unless that is
the primary intended use (maybe it was?)
--
Grumpy Ol' Fred cisin at xenosoft.com