Re: GIMP 3.0.0-RC1

On Thu, 13 Feb 2025, The Natural Philosopher wrote:

On 13/02/2025 04:41, WokieSux282@ud0s4.net wrote:

On 2/12/25 11:01 PM, The Natural Philosopher wrote:

On 12/02/2025 22:54, John Ames wrote:

On Wed, 12 Feb 2025 17:09:18 -0500
"WokieSux282@ud0s4.net" <WokieSux283@ud0s4.net> wrote:
 

As for including size info in arrays ... makes good sense to me. EZ
to know what you're dealing with. Liked the old short-strings in
Turbo - the first byte was the string length.

 There's definitely an argument to be made for including bounds info as
part of the array structure. There's no argument (that I've ever heard)
to be made for making it part of the *type specification.* Any line of
reasoning that says a carton of six eggs and a carton of twelve eggs
are somehow different *kinds* of objects and their contents incomparable
is fundamentally deranged.
 

The problem with languages designed to let stupid people program safely is that as in the case with all highest common factor legislation, the majority suffers to protect the few idiots from themselves.

 >
   But IS THERE ANY OTHER WAY ???
 

Yes.
>

   There have always been some idiots in programming/development.
   That percentage, for a number of reasons, seems to have steeply
   increased.
>
   Almost ALL of western economies absolutely DEPEND on the
   net/cloud/systems in order to function - commerce, banking,
   the infrastructure, transport, energy, supply/demand, mil
   and security - ALL of it.
>
   As it appears very difficult to weed out the idiots, and
   years to create a new class of Competent, the second tier
   approach is to COPE with them. Alas this means much more
   'idiot-proof' computer languages/systems no matter the
   cost/hassle to the competent fraction.
 

You can trade efficiency of the generated code for efficiency in writing it. By adopting standard engineering practicves iof quality control
>
Code needs to be tested and certified like an aircraft or a car.,  and if it doesn't work it needs to go back and be fixed by random code monkeys until it works better.
>
And once you have a good stable design dont fuck with it.
>
Modern software is always being randomly fiddled with to make it more marketable.

There are good examples I think we could learn from. SQlite and Postgres come to mind as software projects that have been quite solid.
I will meet the creator of curl in a few months, and I will ask (if I remember) his opinion on how to write good quality software.

   Nobody wants to hear this, but Real is Real.
>
   As for including type info - limits and more - the effective
   overhead in these days of gigabit flow and GHz multicore chips
   is negligible. As such I'd say to include it one way or another.

>
Well it isn't negligible. My friend who does research into huge mathematical matrices has been busy translating some Intel assembler that makes use of 512 bit registers, into C.
So he can port the code to ARM. It runs at a shade less than half the speed.
Since a full run takes several months, this is significant.
>
He really doesn't need some random memory management getting in the way. His arrays typically exceed the memory size of the machine (128Gbyte I think) and need custom tuning to get swapped in and out efficiently.
>
I think the real problem is that code is written for consumers, even when it needs to be of professional quality. Banking software that ought to be rock solid COBOL is given a pretty face with java and javascript to make the thing appeal to modern numpties who think that a smart phone is 'hi tech' and where it's at..
>
>