,
9 tweets,
2 min read
Read on Twitter
"The publication of scientific Fortran programs" by K. V. Roberts celebrates its 50 year anniversary (doi.org/10.1016/0010-4…) - this is a remarkable and underrated paper which is surprisingly modern and relevant - few gems: (1/9)
- "Programs should be published in source language (rather than binary)"
- "The scientific community should always be able to verify that a published program will produce correct results"
- "A published program will form part of a permanent international literature"
- "The scientific community should always be able to verify that a published program will produce correct results"
- "A published program will form part of a permanent international literature"
- "A program should be fully tested before it is published"
- "A program should be carefully checked by colleagues [...] before it is submitted for publication [...] and it should be refereed"
- "A program should be carefully checked by colleagues [...] before it is submitted for publication [...] and it should be refereed"
- "[programs] should be very carefully documented"
- "Write each program, quite deliberately, for other people to read and understand"
- "Calculations should be as simple as possible to initiate and control"
- "Write each program, quite deliberately, for other people to read and understand"
- "Calculations should be as simple as possible to initiate and control"
- "Define the purpose"
- "Break up a long routine into subdivisions"
- "A program becomes more intelligible if it has a clear modular structure"
- "Break up a long routine into subdivisions"
- "A program becomes more intelligible if it has a clear modular structure"
- "the average program [...] is quite difficult for anybody except the author or his close colleagues to understand or to use"
- "At the present time, it often takes weeks of tedious work to get an existing Fortran program running on a new system"
- "At the present time, it often takes weeks of tedious work to get an existing Fortran program running on a new system"
- "Many programs remain in use much longer than their authors originally expect"
- "scientific software should be controlled by the scientific community, and not by manufacturers"
- "it is important to attract the right people, and one of their main incentives will be the credit"
- "scientific software should be controlled by the scientific community, and not by manufacturers"
- "it is important to attract the right people, and one of their main incentives will be the credit"
50 years later, we are still working towards implementing these suggestions although we have the technical solutions. Publication "in source language" is still not standard. I think one reason is in the last point: "one of their main incentives will be the credit" (8/9)
In the last 50 years we got many new tools but we did not adjust the credit system enough: sharing and reviewing software did not matter enough on an academic CV to those comparing academic CVs. I have hopes that this is now finally changing. (9/9)
