SoylentNews is people
SoylentNews
Over at the Linux Journal, Glyn Moody writes about how the tradition of open science implies a call for the use of Open-Source Software. He notes that even in the first issue of the Royal Society's Philosophical Transactions, which is the oldest scientific journal in continuous publication in the world, the principles for open science are taken as a given. Thus, he explains how open source software, along with both Open Data and Open Access publication, should be a natural fit for scientific work.
When did open source begin? In February 1998, when the term was coined by Christine Peterson? Or in 1989, when Richard Stallman drew up the "subroutinized" GNU GPL? Or perhaps a little earlier, in 1985, when he created the GNU Emacs license? How about on March 6, 1665? On that day, the following paragraph appeared:
Whereas there is nothing more necessary for promoting the improvement of Philosophical Matters, than the communicating to such, as apply their Studies and Endeavours that way, such things as are discovered or put in practise by others; it is therefore thought fit to employ the Press, as the most proper way to gratifie those, whose engagement in such Studies, and delight in the advancement of Learning and profitable Discoveries, doth entitle them to the knowledge of what this Kingdom, or other parts of the World, do, from time to time, afford, as well of the progress of the Studies, Labours, and attempts of the Curious and learned in things of this kind, as of their compleat Discoveries and performances: To the end, that such Productions being clearly and truly communicated, desires after solid and usefull knowledge may be further entertained, ingenious Endeavours and Undertakings cherished, and those, addicted to and conversant in such matters, may be invited and encouraged to search, try, and find out new things, impart their knowledge to one another, and contribute what they can to the Grand design of improving Natural knowledge, and perfecting all Philosophical Arts, and Sciences.
Much more could be done by scientific funding bodies to benefit from Open Source Software, while at the same time many Open Source Software projects could do better in reaching out to the scientific community.
(Score: 2) by fyngyrz on Thursday December 06 2018, @03:17PM (4 children)
The underlying assumption here (and in many similar arguments) is that the openness of an engine means people are in there actually doing the checking.
Generally, that's not the case, at least not just because something is open, and it goes a long way towards invalidating the argument. The one exception tends to be security issues, because the harm that failures cause there tends to be so up-front and obvious after some of it has occurred.
In the case of a calculation error, particularly subtle ones, the fixes come according to both the level of severity and the motivation and the ability to do the fix.
In the case of a commercial operation, the motivation is obvious: money. Damage to a product's reputation, and therefore the reputation of the company that makes said product, comes directly from a perception by the market of "this stuff is broken." This can motivate a considerable expenditure of effort when the problem gets enough publicity. Doesn't mean it will; but it can. They tend to keep the people with the required skillsets around (mind you, I said tend) and so they can point a finger and say "fix that" and be reasonably assured that it will, in fact, get fixed.
In the case of open source, where the motivation for a fix is just to get the fix done, perhaps some small recognition (usually very small indeed) from the community and something to stuff on the resume, the question of whether or not it'll get done is, in its own way, just as open as the question with a commercial operation. Who has the time? Who has the skills? What motivates them to engage with the problem? How will the fix be propagated? How will the users know there's a fix? These things all have to line up, and it's a more complex path than your typical closed-source commercial software operation faces.
There's another issue here as well. For both commercial ops and open source stuff, the more stuff that's out there (and there's more every day), the more technical people paying attention to problems and fixing them are needed. When those resources get stretched thin, fixes (and feature deficits) go without sufficient attention.
In the end, it's just not as simple as "open source is better." Every engine needs fuel. Commercial ops are based around a formal process to get that fuel (money.) Open source ops often don't well define what the fuel is, how to get it, or provide a means to keep it going.
Anecdotally now, I write some open source software. The amount of users I have is pretty reasonable for the kind of stuff I write. The amount of people I have contributing to the code is zero outside of myself. I'd like to be able to say that's because it's feature complete and bug free, but... lol. No. I do try to keep it as best it can be, and perhaps that's simply enough for my users, but I think it's more like "holy shit, there's a lot to this, I've not got the time and/or energy."
(Score: 2) by DannyB on Thursday December 06 2018, @03:56PM (3 children)
But at least someone CAN do the checking. If it becomes necessary. Or suspect.
The lower I set my standards the more accomplishments I have.
(Score: 2) by choose another one on Thursday December 06 2018, @06:32PM (1 child)
The way to check if the science is correct or merely an artifact of dodgy tools is to verify it with a completely different set of tools.
Doing a full code review of an entire tool stack (processor, OS, spreadsheet) each time is ridiculous, unfeasible, and itself prone to error. Reproducing the result on a different stack is none of those things.
(Score: 0) by Anonymous Coward on Thursday December 06 2018, @08:42PM
Nope, I came across an issue SAS and SPSS had the exact same bug in their code becaue both got it from equations published in a retracted paper.
I discovered it because I couldnt get the by hand calculation to work. Then i tried the R calculation, got the same answer as by hand, and was able to look at the code to verify the steps were the same. I will never trust proprietary data analysis tools for more serious problems after that.
(Score: 2) by fyngyrz on Thursday December 06 2018, @06:36PM
Someone can do the checking in a commercial operation as well. They'll even get paid to do it.
Face it; there's no assurance either way, most times, particularly if the original developers have moved on.
There's another thing that happens — instead of fixing existing products, companies make new ones, with brand new bugs, missing things people were previously depending upon, and some fixes.
One commercial example would be Apple; they commonly break things in OS X / MacOS, leave that version of the OS behind, still broken, then release a new version of the OS which may not even be permitted (notice I didn't say "able"... people have hacked it to do so) to run on existing hardware.
One open example would be Qt. Have a Qt 4 project? Qt 4 has lots of bugs. Want to get the fixes the Qt people have implemented? You'll need Qt 5. Qt 4's been outright abandoned. But wait... Qt 5 isn't even close to compatible with Qt 4. Projects that were error- and warning-free will fail horribly. Qt 4 is open source, and just as you say, someone could get in there and fix those problems. But no one has. And Qt is an amazingly popular development environment.
If we're honest with ourselves, we will recognize that the premise of open source's being better by virtue of "it can be fixed" isn't really any more hopeful than the one that says closed source can be fixed. Both are true; both are undependable and situationally relative.
--
Don't anthropomorphize my t-shirt.
It hates that.