[FreeBSD-Announce] FreeBSD 12.0 end-of-life
kremels at kreme.com
Sun May 17 18:46:32 UTC 2020
On 17 May 2020, at 02:48, John Howie <john at thehowies.com> wrote:
> Respectfully, you missed my points, and your counterpoints are the regular industry talking points for why we cannot solve the problem.
They are more than just talking points.
> As for “perfection is the enemy of good”, please tell me you will willingly step into a 737-MAX with its original s/w because it is “good enough”.
It is obviously NOT good enough, which is why the entire fleet was grounded.
And there is such a thing as good enough. Uf you are cutting a piece of wood for a shelf there is a vast difference between 30cm and 30.0000000001 cm.
Exactly (perfect) 30cm is the enemy of good (30cm±0.05).
Now, in some cases, (not cutting a shelf, but other cases) you might need something that is 30.00001±.000005, but that is a much harder goal to achieve, takes many more attempts, a lot more failure, and a exponentially higher cost. And, is a total waste in cases where it is not necessary. Are you designing parts for a NASA mission? Well, then you know why these missions are so expensive.
Perfect is absolute an unobtainable, good is relative and the goal. Things that do not work, are dangerous, are insecure, or cause damage are not "good" (though in some isolated cases, they may be "good enough".
For example, I have a script that I run on my mobile devices that sends the URL of the page I am on to one of my machines. That machine parses the URL and produces a markdown file and a PDF file of the page and puts them into a folder that is synced to my machines. Basically a roll-your-own Instapaper.
It is secure (using key exchange to login to the server) but it is quite exploitable and could cause damage if I sent a malformed URL that exploited a failure in one of the libraries on the system. However, I am exceedingly unlikely to do that, so despite the technical risk inherent in my system, it is definitely "good enough" even if it is not "good" (and certainly not perfect).
> There are plenty of real world examples where perfection should be mandatory, and good is not enough.
Name one thing that is perfect. Just one.
> I may be wrong, but I get the feeling I have been in the software industry longer than you, and maybe it is just my age talking, but we are arrogant in thinking that we can get away with Minimum Viable Product, and (shudder) “fail forward” Agile methodology.
No one is suggesting that.
> Maybe, just maybe, we should take a step back and reconsider what we are trying to achieve.
I prefer living in 2020 with 2020 computers, despite their flaws and missteps, than going back to a "simple" age of 1970s machines that did amazing things, but very limited by today's standards. The cost of modern benefits is vigilance and updating very complex systems.
It is something we are getting better at, and our systems are far more robust AND far more capable. System V (back in the 80s) was relatively trivial to get root rpiveldges on and do anything you wanted to the system, something that happened quite often on the user machines I was logging into at the time. While I have fond memories of those days (mostly involving hunt, forums, rchat, and mTrek), I would never go back to mid 80s computers; nor mid 90s computers, nor pre 2010 computers.
'Everything will be all right. From History's point of view, that is.
There really isn't any other.'
More information about the freebsd-questions