Share
Embed
Software systems and software publications
As given at #AAS229 in the Special Session on sofware organized by Alice Allen (ASCL).
David W Hogg
January 06, 2017
Transcript
20 slides
-
relationships between software publications and software systems David W Hogg
(NYU) (MPIA) (Flatiron) twitter:@davidwhogg GitHub:davidwhogg -
example 3x3 quadratic-fit method for centroiding stars in the SDSS
photometric pipeline Lupton et al, 2000-ish; Vakili & Hogg, arXiv:1610.05873 -
three related publishable entities a software system or pipeline a
publication primarily about that system (pure software publication) a “science” paper that also describes the system (paper with a software component) -
citation of software cite the source code (eg, via ASCL)
cite a pure software publication cite a paper with a software component -
benefits of writing a pure software publication serves as discoverable
documentation preserves ideas in the code legitimizes the code; distinguishes it from competitors gets you refereed citations for your software work -
example our emcee paper (2013 PASP 125, 306) has >1000
citations and is getting 40 per month -
costs of writing a pure software publication takes time and
money produces a publication that some don’t respect distracts from other scientific or career goals (this is an opportunity cost) -
time scales different costs and benefits accrue on different time
scales eg, the time cost and the documentation benefit accrue immediately eg, the preservation of ideas benefit and the opportunity cost accrue over very long time scales -
#LTFDFCF how you weigh costs and benefits depends strongly on
your career stage and also the importance (to yourself) of your software work -
random thought: preservation of software preservation of installable executable preservation
of source code preservation of fundamental ideas -
the world is changing respect for pure software contributions is
much, much higher than it was even 5 years ago (let alone 10 or 20) (Thank you, AAS community: Some of my students bet their careers on this.) -
benefits of writing a paper with a software component preserves
ideas in the code, or some of them legitimizes the code; distinguishes it from competitors refereed citations for your software work situates the software in an appropriate scientific context -
appropriate scientific context (I know this isn’t a talk about
open-source software, but:) one of the biggest costs of releasing a software system is time spent batting down inappropriate uses -
costs of writing a paper with a software component takes
time and money (probably) doesn’t serve the documentation value of a pure software publication -
double jeopardy I don’t like the idea of refereeing the
software systems in addition to the papers about them. I don’t like the statistical editor comments on data analysis papers submitted to the AAS Journals. -
benefits of having no traditional publication save time and money;
no refereeing headaches either citation of the software is direct citation of the relevant work the ASCL and ADS and GitHub have made all this very easy -
costs of having no traditional publication documentation remains necessary citations
are (currently) unrefereed (and less respected, perhaps) uncertain long-term value of citations; preservation issues hard to distinguish your software system from competitors -
advice when you release software, explicitly state how you want
it cited, alongside or very near the license statement -
conclusions for me, the benefits of producing traditional publications are
legion (though sometimes a bit intangible) both pure software publications and papers with a software component are very valuable the (new) direct cite-ability of software does not much diminish these benefits