David Delbecq wrote: > En l'instant précis du 14/12/07 06:20, Mark Wedel s'exprimait en ces > termes: >> Fair enough. But suppose someone checks something in using adobe photoshop >> format, and a month later I want to change the image, but don't have >> photoshop. >> >> Is it OK then for me to remove the original (source) photoshop image and >> just >> have a PNG? Even if the original author is still around, maybe he doesn't >> have >> time to make the change in the near future. >> > Well, you can't open the photoshop file, you remake a file using, > example, gimp and commit png, gimp, remove photoshop. > The source is now gimp. If the final file is similar to original one and > original author wants to restart from his photoshop file, fair engouh. > He retrieve it from the svn history, manipulate it and commit this: new > png, removed gimp, new photoshop. As i said, to the opposite of cvs, svn > does handle properly deletions. Original file is still available in history.
Ok - I have no problem with that being the method used. People will just need to make sure they svn delete the source file - otherwise there is real risk of someone seeing the source file and overwriting the modified png. OTOH, the changes lost are those that the folks who wrote the png made, so there is real incentive for them to make sure they remove that source file - it is their changes being lost, not someone else. >> And gimp's .xcf files are another odd case here. Pretty much every linux >> distro will have gimp, but folks using windows or mac would need to download >> it. >> Should it be considered proper that the mac/windows person download gimp >> and >> update the .xcf file, or is it ok for them to just write out a png and >> remove >> the .xcf source? >> > It's probably less work for them to download the xcf / edit / save as > png then open png in photoshop, extract elements, remakes some layers, > save both. Not sure I understand your explanation. Wouldn't it be less work for them to just open the PNG in photoshop, and not deal with the xcf files at all? > I think the main question is "do we comply to GPL to the point we must > give *all* sources, including those of our png? If yes, then if someone > modify something with an obscure program, he must commit the source too :/ The meaning of a 'source' for an image is somewhat misleading. For one, it is hard to prove that there even is a source (someone could be using some basic image manipulation program that just reads and writes pngs and doesn't have any meaningful native format). This has been the case - certainly when mass changes have been done (like xbm -> xpm -> png conversion, as well as 24x24->32x32 conversion, things like netpbm have been used). But I also use the term meaningful format here, because I think that is relevant. There are lots of good examples why having xcf data could be useful - if the layers are separate, can manipulate those to different images, etc. However, there are also lots of cases where having the xcf image may really not be useful. For example, if I take an image and it needs some cosmetic changes, I'm not likely to do that on a layer, I'm likely to do that on the base image itself. All the xcf image would get the next developer is potential to undo the changes from within gimp (I think gimp saves the undo history, but not sure). Is it really worthwhile then to have the xcf then? I'd say probably not. _______________________________________________ crossfire mailing list crossfire@metalforge.org http://mailman.metalforge.org/mailman/listinfo/crossfire