|
On 12/3/18 8:48 AM, Ankur Sinha wrote:
On Mon, Dec 03, 2018 12:41:41 +0100, Kevin Kofler wrote:Since correctness is really important here, if upstream does not test the toolboxes against Octave, we shouldn't eitherI think that if it runs at all, we should ship it. Some upstreams seem pretty conservative. E.g., SPM seems to have done a lot of work on Octave compatibility already, and the page seems to imply that it will more or less work, with some issues, they just do not support it still. Fedora, in contrast, is a forward-looking distribution that is about shipping Features First and with Freedom (i.e., no MATLAB!) included.No, I disagree with this. I assume that the package you propose would work with Octave and Matlab co-installed. If it didn't, I would think it would be unacceptable, because it would effectively be evicting Octave. It seems to me that there has to be some sort of O.vs.M. configuration, so people who don't want to take the risk can still use the Matlab setup. We are not making life harder for those folks: they have to manually install Matlab regardless whether the package Requires Octave or not.
These are not user-end applications where an annoying bug may be fixed in a future release and make everyone happy. These are toolboxes that are used for analyses of data, and the results from the analyses contribute to science. Then, other studies will build on these results, and so on. If there's any hint of lack of correctness, being conservative makes sense---it is too much of a risk. A minor issue can undo years of work and progress. It would be very bad if an issue in our provision of the toolbox with Octave resulted in a retraction, later on, for example. [...] If a researcher uses a MATLab toolbox with Octave, it is their responsibility to check for correctness. If we provide them, it is ours, and this is not a responsibility we have the man power to take on. We'd rather rely on upstream. Software licenses usually claim that "the customer is solely
responsible for selecting the software and determining the
software’s suitability for the customer’s particular purpose". I
would not be surprised if Matlab license contained a similar
clause, and the researchers should check for correctness anyway. > I think that if it runs at all, we should ship it. It seems to me that 'runs at all' is too low of a bar; I would
personally prefer that it passes some regression tests. Providing
such tests, which presumably would also allow people to
sanity-check against both back ends, would be a nice added value
for both end users and for Fedora. |
_______________________________________________ devel mailing list -- devel@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxxxxxxxx Fedora Code of Conduct: https://getfedora.org/code-of-conduct.html List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@xxxxxxxxxxxxxxxxxxxxxxx
