Letters: August 2020

Last month’s article on the coding crisis in computational chemistry is part of a much wider problem with software being in error or completely impenetrable. I thought it might be useful for members to share some experiences, maybe anonymously, so here is my two penn’orth.

Firstly, while studying x-ray crystallography as part of my PhD in the 1970s, I learned that the widely-used diffraction pattern computation had been coded by someone called Sheldrake (I think) years previous and had been much modified. However, several regions of the code were ‘no-go areas’ that were not to be modified because no-one understood how they worked and changes had unpredictable effects.

Secondly, in my working life years later we had a tried-and-tested – i.e. old – software-based warehouse inventory management system. In a routine, monthly, perpetual-inventory on 10% of stock, we discovered the computer system automatically changed anything in quarantine status back to free-sale status! This monumental error had eluded decades of use of the software in multiple locations and had any of the quarantine material then been shipped, multiple recalls could have been needed.

At the time the warehouse software problem was discovered, we were just getting serious about computer system validation and frankly, you can see why.

Anonymous correspondent