I have an opportunity to do some hacking this weekend and the next, which should (frickin’ finally) yield some new public versions of each lib. Each should have more than one fix present, particularly in light of the really solid public bug patches I’ve received over email (thank you!).
Today, I found out the hard way that DD_roundies (and by inference, I suspect DD_belatedPNG as well) really hates HTTPS traffic and will cause ‘operation aborted’ errors in some cases (even though I’m NOT doing an appendChild to an unfinished element). I conjured a solution, nonsensical as it may be… I suppose that is fitting, given that the PNG lib in particular is nonsense to begin with.
Anyway, I’m looking for some feedback, on one idea in particular: both libs require the silent addition of invisible image elements to a given page to determine the sizes of all images that get applied to VML elements. I, and pretty much anyone who has used DD_belatedPNG to a great degree, have noticed significant initial page load lagging when using that lib. I suspect it is in part because of the addition of the above mentioned “size-finding” images.
How would you feel about the lib requiring you to add an object literal that would inform the lib of how big the image sizes are… sort of like a configuration object?
This would enable me to skip the whole size-finding bit. However, it sounds like a pain in the butt. A compromise: if no config object is detected, the lib runs the size-finding bit once, and spits out a config object in form of some javascript code that can be copied out of a textarea.
Even if it doesn’t save much CPU time, it would at least prevent the need to download a given image twice just to display it once.
