On Thu, May 26, 2016 at 02:17:11PM +0100, José Abílio Matos wrote: > On Wednesday, May 25, 2016 1:39:56 PM WEST Scott Kostyshak wrote: > > I'm still curious someone has thoughts on the above idea. The main > > question I have is do we usually know when we are not producing > > equivalent documents (and thus it is easy to issue a warning)? I assumed > > so but I don't have experience with this. > > > > Scott > > lyx2lyx can emit warnings but we do not process them in lyx. So for the > moment > that will not work. If the code follows the usual conventions the warning > messages are easy to catch and process.
I see. > We should put things in perspective, for most of the features the convertion > is seamless. The backport is an added bonus that allows users to read > documents from the next version. I don't think that the users understand that it is an added bonus and might not work in many cases and that we knowingly allow for data loss. On the other hand, I do not remember a single bug report from a user complaining about exporting to an older version so perhaps my concern is indeed not worth it. > Ideally this should a last resort scheme > because we do not test for back and forth convertions to guarantee a stable > cycle. There has been some progress towards such tests, but the tests are not unit tests. > IMHO the ROI return of investment for this is so small that it is not > worth, that is the amount of work required is so large for a so small benefit. > > And just as in the case of the question of qt4 vs qt5 as long as the code to > convert back covers all the (relevant) cases we fulfill our promisses. :-) > > Again I reiterate that this is my humble opinion. :-) Makes sense, thanks for the explanation. Scott
signature.asc
Description: PGP signature
