[FRIAM] Meanwhile, back on the troll farms

[thompnickson2 at gmail.com]

Marcus, 

 

Thanks for taking my question seriously.  I understood what I was talking about even less than I usually do. 

 

Let’s say I was an evil genius and wanted to introduce evil code into a project on github.  What would happen?

 

N

 

Nicholas Thompson

Emeritus Professor of Ethology and Psychology

Clark University

 <mailto:ThompNickSon2 at gmail.com> ThompNickSon2 at gmail.com

 <https://wordpress.clarku.edu/nthompson/> https://wordpress.clarku.edu/nthompson/

 

 

From: Friam <friam-bounces at redfish.com> On Behalf Of Marcus Daniels
Sent: Thursday, May 7, 2020 11:05 AM
To: The Friday Morning Applied Complexity Coffee Group <Friam at redfish.com>
Subject: Re: [FRIAM] Meanwhile, back on the troll farms

 

Nick writes:

 

< What exactly IS the policing mechanism in open source.  Darwinian? Reputational?  Does this HAVE to provoke a crisis of confidence in the general public?  Or could it be seen as a heroic thrown-together first step that is now being improved? >

 

They are whining about simple or absent unit tests as a litmus test for whether the code is reliable.   It’s like saying you don’t dare drive your car if you didn’t take out its alternator and test its voltage output last week.   ‘cause someone might have changed the alternator!   Eventually there will be consequences if the alternator fails, like stalling or the battery dying.   Same thing in a big simulation.   All of the parts and pieces of a simulation are there for a reason and global things will start to change in noticeable ways if something is broken.   I would say getting mechanisms working correctly is less difficult that choosing what mechanisms are appropriate in the first place.   Usually in use of a simulation one has instrumentation available on almost everything, and there is a constant checking and double- checking even if those checks are not embodied in automated tests.  Automated tests can even give a false sense of security, because they may not deal with the parameter ranges that happen in with the coupled system.  If you would rather have a bunch of unit tests, or to have modelers using and stressing the code every day, you have the wrong priorities.

 

My irritation is with the notion of unit tests as a prerequisite for code reliability.   There are tighter ways to integrate assertions of code behavior with the code.   The bandwagon obsession with unit tests is in some sense an obstacle even better practices.   I wouldn’t even call them trolls, because a troll has intention to rile people up.  These folks are more like pompous ditto heads who feel the need to posture about the right way to do software engineering.   People that love unit tests love not understanding the problem they are solving, and prefer to work in pieces.   This take a is a little harsh, but in this context (advising COVID-19 policy) I don’t find the behavior very helpful.

 

Marcus

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://redfish.com/pipermail/friam_redfish.com/attachments/20200507/6130cd08/attachment.html>