Something I dealt with a long time ago that has become a sort of rule for me, regardless of how true it might be: Scientists, and University researchers (the tenured ones) hate learning new programming languages and methods. There’s decently good reason behind it, as far as I can tell. I used to support an archive of weather satellite data. Whenever we had a software stack upgrade come in, the scientists grumbled because it meant they had to revalidate large swaths of their data with the new versions to make sure all results were reproducible. One thing they never did, if they could help it, was change the base code they used to generate those results. That would mean much more work. Also, if they wanted to to come up with a new subset of the data, they wrote it in what they knew. Usually Ada! Supporting this is how I learned that we couldn’t get an Ada compiler that would produce 64-bit binaries. The compiler binary itself was 64-bit, but that was it. From what I could learn, SGI had produced a 64-bit compiler for IRIX (I think - which ironically we were migrating from to x64 Linux clusters), and PGI gave up on theirs for “lack of consumer interest.”
Something I dealt with a long time ago that has become a sort of rule for me, regardless of how true it might be: Scientists, and University researchers (the tenured ones) hate learning new programming languages and methods. There’s decently good reason behind it, as far as I can tell. I used to support an archive of weather satellite data. Whenever we had a software stack upgrade come in, the scientists grumbled because it meant they had to revalidate large swaths of their data with the new versions to make sure all results were reproducible. One thing they never did, if they could help it, was change the base code they used to generate those results. That would mean much more work. Also, if they wanted to to come up with a new subset of the data, they wrote it in what they knew. Usually Ada! Supporting this is how I learned that we couldn’t get an Ada compiler that would produce 64-bit binaries. The compiler binary itself was 64-bit, but that was it. From what I could learn, SGI had produced a 64-bit compiler for IRIX (I think - which ironically we were migrating from to x64 Linux clusters), and PGI gave up on theirs for “lack of consumer interest.”