what came first? notebooks in mathematica or knuth's ideas on literate programming[1] ?
regarding notebooks themselves, i feel like they're a high concept idea but i've yet to see them really click for me in practice. i find the small cells for code to be extremely unergonomic and that the interspersal of code and plots to be distracting from both the code and the plots (although pretty fantastic for demonstrating high level features of a library, programming language or environment).
on a more fundamental level, i completely agree that mathematical notation is lossy, and that it takes a lot of skill to go from some arcane notation to an actual sense of what the relationships are- but, it requires no specific functioning technology to do so. i can review a paper from 100 years ago and understand it, where running a computer program from 20 years ago can be a challenge at best.
i think that additional high touch experiences for data exploration and teaching are fantastic ideas, but i also think that maybe the base level of communication should be kept simple; both for the purposes of maintaining accessibility and history. where the linux kernel developers insist on 78 column listservs, maybe scientists should insist on camera ready documents when it comes time to share.
i think that everyone agrees that better science would come from full data and code being supplied with publications, but interop is quite difficult as-is keeping code alive. i suppose the big question is: does it make sense to move science towards how software is done, where every bit of code is actively maintained over the years to avoid code rot, or does it make sense to come up with a scheme of freezing and archiving computing environments used in science so those in the future may be able to reproduce results or errors as they see fit. (something like, every paper must ship with a vm image for a widely available architecture that includes no proprietary code and all data used for results)
interesting questions. how to fundamentally change scientific communication such that it is enriched with data and code properly is a harder/organizational problem that i think many have tried to solve (not to mention how this ties into another problem in science- idea validation/replication and knowledge rot). building software systems for exploration and data analysis (ie; computer as partner in exploration) sounds much more fun and likely to produce useful results!
regarding notebooks themselves, i feel like they're a high concept idea but i've yet to see them really click for me in practice. i find the small cells for code to be extremely unergonomic and that the interspersal of code and plots to be distracting from both the code and the plots (although pretty fantastic for demonstrating high level features of a library, programming language or environment).
on a more fundamental level, i completely agree that mathematical notation is lossy, and that it takes a lot of skill to go from some arcane notation to an actual sense of what the relationships are- but, it requires no specific functioning technology to do so. i can review a paper from 100 years ago and understand it, where running a computer program from 20 years ago can be a challenge at best.
i think that additional high touch experiences for data exploration and teaching are fantastic ideas, but i also think that maybe the base level of communication should be kept simple; both for the purposes of maintaining accessibility and history. where the linux kernel developers insist on 78 column listservs, maybe scientists should insist on camera ready documents when it comes time to share.
i think that everyone agrees that better science would come from full data and code being supplied with publications, but interop is quite difficult as-is keeping code alive. i suppose the big question is: does it make sense to move science towards how software is done, where every bit of code is actively maintained over the years to avoid code rot, or does it make sense to come up with a scheme of freezing and archiving computing environments used in science so those in the future may be able to reproduce results or errors as they see fit. (something like, every paper must ship with a vm image for a widely available architecture that includes no proprietary code and all data used for results)
interesting questions. how to fundamentally change scientific communication such that it is enriched with data and code properly is a harder/organizational problem that i think many have tried to solve (not to mention how this ties into another problem in science- idea validation/replication and knowledge rot). building software systems for exploration and data analysis (ie; computer as partner in exploration) sounds much more fun and likely to produce useful results!
[1] https://en.wikipedia.org/wiki/Literate_programming