On Hën , 2003-10-27 at 08:38, Jeff Waugh wrote: > <quote who="Mark Finlay"> > > > The reason we choose to store metadata for images is mostly to do with > > the fact that for any given image only one or two of the scaling types > > make sense, and for translucent images it's useful to have a background > > color/background type associated with it. > > > > This will have two major benefits IMHO: > > 1. For the default wallpapers we ship with Gnome(/Ximian/Redhat etc...) > > we can choose the scaling type, and for the transparencies we ship, we > > can choose nice background colors to go with them. > > OS X autodetects which scaling method to use and allows you to override it Yes. And this will too. However, there are two code paths one can take. 1) Read the preferred scaling type from the XML data file. 2) Do a lot of math and load the image into a pixbuf so you can get at the necessary data, and possibly choose the wrong scaling type. We can do the latter every time, but it will cause slowdowns and may just end up causing the user to get frustrated when the wrong type is chosen and they constantly have to change it. > if you wish. Colours and transparency are an added bonus, but I honestly > didn't know this worked until late 2.1.x. Probably something to relegate to > cool distro and power user tricks. Colors are not an added bonus. Colors have been a part of Windows (that other OS that we are trying to migrate people from), as well as MacOS for many many years. > I'd question the merit or usefulness of saving these, however. Seems like > overkill for something that could be handled dynamically without extra work Overkill would be to calculate it every time, rather than doing what the user told you to do. However, doing the later, is much more useful than not. > for the user. Same goes for distro settings - they'll be shipping a preset > wallpaper, and probably providing a few others just for fun. Is there a > great need to store super-duper-amazing metadata to make a background image > do the right thing when a user clicks on it? So that we do do the right thing. If we don't store the data, we will end up having to re-calculate everything, for every image, when it's loaded. That is a much higher drain on the system resources than reading a string from a file. > > 2. When a user adds an image/transparency they only have to set the > > scaling/background colour once. > > I guess I'm not convinced that this is so deeply troubling that we need to > apply clever magic to it. ;-) Needs shouldn't be the only thing to warrant features and/or magic. Many people want it. And it's not magic. It does exactly what I expect it and wrote it to do. And it works very well. -- dobey
Attachment:
signature.asc
Description: This is a digitally signed message part