On Friday, 5 January 2018 at 13:22:00 UTC, Paolo Invernizzi wrote:
- be quantitative: your download statistics are a good start,
try to collect from commercials statistics about the length of
the codebase, the compilation times, how many are using a
feature (C++ integration, allocators, scope when polished).
- be fact driven: analyse your own predictions about metrics
with what you are as results from measuring, and iterate on the
next decisions (also) based on that."
/Paolo
Yes, quantitative information is always good for making sense of
things ;-)
I like how linux kernel development do their reports .. not that
D is should be compared to the scale of the linux kernel
development effort .. but still .. numbers, tables, graphs ..
they provide a nice high level overview .. that you can't get
when you're stuck in the trenches.
btw. interesting fact.. in a 10 year period, the linux kernel has
gone from 8 million lines, to almost 25 million lines. If they
average 0.5 defect per KLOC (a very convervative estimate), than
means that 90 percent of the public cloud workload, 62 percent of
the embedded market share, 99 percent of the supercomputer market
share, 82 percent of the world’s smartphones, and nine of the top
ten public clouds... are all running a operating system kernel
with around 12500+ bugs in it.
Jeepers!
