Yikes! Yes, converting our PNGs to "indexed" is a very bad idea - the conversion, like any conversion of an image into a 8-bit color space, is generally lossy. Indexed PNGs, or PNG-8s, by their other name, are 8-bit by definition, and many of our images, especially mine or wayfarer's, use more than an 8-bit color space. Converting down to this typically shifts all of the colors in the image, slightly. In very basic principle, it might have conferred some space savings, but in practice, that is really only through discarding data, since the folks who made the PNG-24 spec did a pretty good job.

Also, various software programs have niggling incompatibilities with png-8s.

I've learned this only through experience - some of the images that I had been replacing for the team-color commit, I had made as png-8's back in the day (c.f. human royal guard line). These I had to jump through some serious hoops before I could import them into photoshop (essentially, photoshop had de-linked and lost the mask, and I had to export the file to a tiff via an alternate program, which recovered the masking information, then import into photoshop and re-apply the mask data, and then export the thing to a png-24). Evil, let me tell ya.


Anyways, no hard feelings about goofing this up - I've certainly had my fair share.
Frame's suggested fix is the way to go.

On Mar 14, 2006, at 3:42 AM, Hogne Håskjold wrote:

Hi

Torangan: Your PNG optimization has changed a lot more images than I first thought. _All_ units that has been changed to indexed and back have less saturation now and look far worse, and in addition they no longer match the !changed pngs. This is becoming a huge mess, and I think we should revert the entire opt. commit and all subsequent attempt to fix it again. Since you got on in this mess to begin with I think it's your responsibility to sort this out.

--
mvh                 (o_
Hogne Håskjold      //\
                    V_/_


_______________________________________________
Wesnoth-dev mailing list
[email protected]
https://mail.gna.org/listinfo/wesnoth-dev


_______________________________________________
Wesnoth-dev mailing list
[email protected]
https://mail.gna.org/listinfo/wesnoth-dev

Reply via email to