Timeline for Can I trust a paper that hides implementation details?
Current License: CC BY-SA 3.0
9 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Dec 7, 2015 at 23:55 | comment | added | Stuart Golodetz | @DaveRose: For the record, I'm not arguing that their code should be hard to build -- I spend a lot of time trying to help people improve their build processes -- I'm merely arguing that it sometimes is. That may be because they haven't bothered to set up a decent build process, or because they don't know a better way of doing it. The underlying problem in both cases is that building things reliably, particular things with lots of dependencies, is not easy. (That comment's based on a fair bit of time spent automating builds of third-party code.) | |
| Dec 7, 2015 at 23:51 | comment | added | Stuart Golodetz | @DaveRose: I was primarily arguing that your assertion that "it is not hard to compile any mainstream languages" is by no means always true. In general, setting up cross-platform builds that work reliably on all setups (including those to which you may not have access) is not trivial, even if you have a reasonable amount of experience. Many researchers (particularly early-stage ones) don't have that experience, and their code can be hard to build as a result. | |
| Dec 7, 2015 at 20:57 | comment | added | Jouni Sirén | Most computer scientists can't write quality software, because they don't have the training or the experience for it. Even worse, research code is usually written by the least experienced members of the team. By the time people start learning best practices, they usually advance to higher-level stuff, and the programming tasks are assigned to new students. | |
| S Dec 6, 2015 at 23:27 | history | suggested | darthbith | CC BY-SA 3.0 |
Correct grammar
|
| Dec 6, 2015 at 23:26 | review | Suggested edits | |||
| S Dec 6, 2015 at 23:27 | |||||
| Dec 6, 2015 at 23:26 | comment | added | o-0 | @StuartGolodetz Also... look at big main stream languages/operating system that are open sourced which you can build them up from the source with all their dependencies. In my opinion (that might be extreme to some academics), if the code can not be run or demonstrated, then the researchers did not their job. If they want to publish a paper on just theory then thats another ball game. | |
| Dec 6, 2015 at 23:21 | comment | added | o-0 | @StuartGolodetz I know that, however this should not be an excuse for the author of the paper to provide a crappy code that can not be run. They can easily contain all their code and all their dependencies into one zip and write a bash for it or to use third-party containers. Again, this is a problem in CS domain and some people unfortunately take an advantage of it. This is a grey area that you need to take a decision, for the faith of the paper, with a senior lecturer. | |
| Dec 6, 2015 at 22:13 | comment | added | Stuart Golodetz | It can be extremely hard to build code, depending on things like the language, what dependencies are needed, and how the build has been set up. Research code is actually often quite hard to build, because little effort is put into setting the build up properly (often it's only easy to build on the machine of the original researcher). | |
| Dec 6, 2015 at 21:01 | history | answered | o-0 | CC BY-SA 3.0 |