On Thursday, September 17, 2020 3:07:23 P.M. CDT Thomas Goirand wrote: > On 9/16/20 2:55 PM, Steven Robbins wrote: > > Since you're soliciting opinions, here's mine. In the absence of a > > documented consensus, ftpmaster should respect the packager's judgement > > rather than reject on their own personal opinion. > > Reviewing the packaging is also part of the FTP master job. The debate is over the desirable _extent_ of such a review. I would also say that a review that offers suggestions for improvement is totally welcome. On the other hand, an opinion masquerading as an unstated requirement to pass the gate is an irritant. I like this discussion on the Tone of the Review https://stackoverflow.blog/2019/09/30/how-to-make-good-code-reviews-better/ > On 9/16/20 2:55 PM, Steven Robbins wrote: > > Thorsten's observation ("... is much too large") is completely > > arbitrary. Also, why does size matter? If the files are necessary, > > they will show up somewhere. Why do we care which tarball they are > > part of? > > The above shows you haven't understood what the problem is. That is certainly possible! > I replied already to Andreas, though here's my thinking, hopefully that > was the one of TA as well. > > With a separate source package holding the data, the data set, > typically, will be uploaded *once*, then we may see new revision of a > tiny debian tarball. I also don't think such a package will need so many > revisions anyways. > > On the other hand, the package which needs to be tested with this > dataset may need to be often upgraded to the latest upstream revision. > Uploading a huge debian tarball each time isn't optimized. You are right. I don't understand the problem. If the size of upload is a concern, surely the uploader is the most invested in such optimization and would be a good judge of that. Why do we need ftpmaster to second-guess it? Now, I'm a disinterested observer and have not looked at either the data or the code in question. It may well be true in this case that the test data rarely changes. However, my experience at work is that we update test data not infrequently so I think I'd be careful of drawing conclusions on what is "typical". Regards, -Steve
Description: This is a digitally signed message part.