I've got a major question...

[Steve O'Hara-Smith]

On Thu, 27 Jun 2019 08:01:34 -0400
Jerry <jerry at seibercom.net> wrote:

> On Thu, 27 Jun 2019 11:37:12 +0100, Steve O'Hara-Smith stated:
> >On Thu, 27 Jun 2019 11:03:40 +0200
> >Polytropon <freebsd at edvax.de> wrote:
> >
> >> This is especially true when you need software for a special
> >> purpose. Maybe I would call them "niche software", but it is
> >> actually in heavy use, like accounting, payroll, time-work
> >> management, or ticketing systems.  
> >
> >	Many many years ago the approach to choosing a computer system
> >	went
> >like this:
> >
> >1: Establish your requirements
> >2: Find or commission software that best supports your requirements
> >3: Select a platform that runs the software you selected
> >4: Iterate round 2 and 3 to optimise cost/benefit
> >
> >	These days the common approach seems to be:
> >
> >1: Select platform by default
> >2: Accept software on offer
> >3: Adjust requirements and procedures to suit software
> >4: Bitch when you find yourself locked in
> 
> I agree with all except #3 in the second set.
> 
> 	sed 's/software/operating system/'

	Nope it's the software you accepted because it was on offer
with/for the OS you got by default that means you have to adjust your
procedures.

> #4 is right on no matter what system you are utilizing.

	Avoiding lock-in is harder and requires thought and a policy of
doing so from the start and accepting any limitations it imposes[2]. One
implication is that all data must be stored in documented formats[1] with
multiple implementations of tools available. Way back when I was learning
how to design electronic things or production one principle that was
strongly held was that nothing should be single-sourced, it's a good
principle IMHO.

[1] Openly documented isn't required provided you can obtain the
documentation and the right to commission an implementation.
[2] They're not really onerous but do require a willingness to not go with
the flow.

-- 
Steve O'Hara-Smith <steve at sohara.org>