Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 14 Feb 2018 23:08:26 -0800
From:      C Gray <frankfenderbender@council124.org>
To:        freebsd-questions@freebsd.org
Subject:   Re: any problem going from 9.x (don't laugh) to 11 directly?
Message-ID:  <8212BAA6-DC01-4AD5-BCBF-012D97447060@council124.org>
In-Reply-To: <7795e899-160e-f6af-c02d-6fa44982f864@ShaneWare.Biz>
References:  <20180215011907.6620E1B2DE28@ary.qy> <868tbvhwix.fsf@red.stonehenge.com> <7795e899-160e-f6af-c02d-6fa44982f864@ShaneWare.Biz>

next in thread | previous in thread | raw e-mail | index | archive | help
Moving to something called stable means that the build (not the release) =
is stabilized.=20
At Network General, Tektronix (Printing, Test&Measurement, and Graphics =
Workstations divisions), Intel, Cypress Semiconductor (EDA), Analogy =
(EDA), and Mentor Graphics (EDA), they all had development use verified =
tools and modules (rather than the newest ones) waited until a major =
release had 2-4 minor releases which fixed problems that users would end =
up debugging to make up for incomplete pre-release testing policies. =
[Note: new car models have more problems & glitches than ones =
recalled/fixed or return-model releases; it goes for anything =
manufactured in a world where fixing costs and the "customer is always =
right" has been mutated into "the customer is always ripe". ]=20

In Test and QA (integrated in with Development) we were taught (by =
losing customers and costs/time/resources lost to fixing issues directly =
related to poor-planning, short-cutting, and thinly-testing) NOT to pass =
those inherent problems on to later and to end-users/customers, because =
it all comes back to you in the end... and esp. if the code was =
third-party or partner software which is still "green", i.e., untested =
by real-world users doing more real-world things than most =
release-builders have time or look-ahead imagination to test. It may be =
reaching a "not supported any more " status, however, that often means =
that it does need as much support as a major release. So, mixed in with =
dropping support way too soon, an almost-fetishized practice (of =
uncaring and 'could-care-less'ness learned from the profitable =
bad-examples of Microsloth and [cr]Apple, people now must own the =
fall-apart software, tools, etc. which that lackluster work ethic =
laziness yields.

Spaghetti code, non-modularity, platform-dependencies... they're all =
coming back as "computer science" is fully replaced by the rapid-release =
strategies of  "computer marketing". Planned obsolescence is a =
behavioral issue more than a materials problem, and as a view accepted =
tends to hurt you directly at a cellular level rather than indirectly at =
an update level. Tweaking ought to be left up to tweakers. Code should =
be better-[PRE]tested. We QA folks were usually hated by sales and =
marketing types, and at first by developers, however, before we were =
excommunicated, they actually saw that we were code-friendly and =
lengthened weekends without the tension and fret caused by =
back[ground/mind]ed debugging.

Think hard about going "new", because the processes which brought all of =
the past bugs are still creating new-and-improved ones unless, of =
course, those processes are considered part of the product, and their =
improvements and need-for-improvement are both heeded, implemented, =
well-documented, and verified as solved, which, the newer releases of =
practically all software these days are de-evolving back to a =
pre-Knuth/pre-Djikstra era.

If FreeBSD r11 doesn't show me less issues than r10 which should have =
less issues than r9, then the opposite may actually be true (or TrueOS).=20=

An absence of process improvement -- which must invariably also include =
and demonstrate an improvement of itself -- will only make more problems =
worse.
As timed-releases of flaws become "newer" and "newer", the time-tested =
fixing-of-flaws become older faster, approaching non-existence in the =
calculus of release policy.

The drive to eradicate real-world stability is lost to the fetish of =
"new"-ness.=20
That is a capitulation to the extraction and exploitation of time.
As Orwell put it quite well, and which also covers the speeding of "now" =
into a disappearing "then", the "future" really ceases to exist at all:
	"Who controls the past controls the future. Who controls the =
present controls the past."
The squeeze between deleting the past by PCers and PCs 'transmerging =
humans' in the future is suffocating and the only way to stop to allow =
[an unsuffocated] thinking is to do just that....=20

My suggestion would be to go for 10.4! It's tested more to itself than =
11 is to itself. Other peoples release habits have little to do with =
actual quality because they do not assure QA. Time allows people to use =
"the town square" approach to solving issues, and at least creates known =
"avoid" flags and "workaround" shares. Even if r11 will be better, let =
it demonstrate it first. There's not a shortage on upgrading and why do =
it all if the time resuscitating increases in its breadth and/or depth. =
Maybe check the areas where you'll most use 9 v. 10.4 v. 11, if tickets =
open-closed in those areas are indeed tracked, so you can see if an =
intended high-use module/function has had any changes to it and too what =
degree of change. If that information is missing in the build-release =
process, then the build-release process needs to be transcended as it =
now exists? Bugs need to be prevented... not discovered.

Maybe trust your experience rather than a release#-incrementing =
algorithm, esp. when it is followed by ".0".

Decide based on the rate of flaws rather than on a speed of releases, =
and consider that "change for change's sake" is a pastime (the passing =
of time) and investing habit of wasting your future time by "selling the =
past short".

best,
chris




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?8212BAA6-DC01-4AD5-BCBF-012D97447060>