1. I mostly use PNG and AI files for my images, all imported by reference and with a border on the AF set by an Object Style. I've found that nearly all of my PNGs lose their borders when I publish to HTML. The odd one or two keep their borders, but I can't determine why not. On the contrary, all of my AI images keep their borders in the HTML.
Anyone know why this happens?
My workaround for this is to place a very tiny--and blank white--AI image behind the PNG, in the same AF. Dunno why this works, but it always does.
I've tried using the STS to place borders around images, and it works great, but it treats equations as if they were images, so they get borders too. Our style standards say no borders around equations. Argh.
2. Some of the images lose only one or two of their borders. I've managed to fix this by enlarging the AFs so there is more white space around the image. If that fails, I move the border from the AF to around the image inside the AF.
I've figured out that the border width of only 0.5 is part of the problem...if I increase the width to 1.0 the missing borders show up when I republish. That doesn't explain why 0.5 is good enough for some images but not others, though...
In FM, borders on PNG raster objects (and probably other raster formats) are not (and can not be) an actual border attribute on the imported graphic object itself, because PNG doesn't encode a concept of borders. If it did, border sizes smaller than one pixel might be expected to be a problem.
The border on a PNG is instead applied by overlaying a separate border object set, which is somewhat surprisingly not even an empty rectangle. It is four separate thin rectangles, one for each edge, each having a fill of your chosen border color, and no border of their own. They are overlaid on each other at the corners (not even mitered). They aren't even strokes, when deconstructed in Illustrator (which is often a great way to discover what FM is actually doing).
As vector polygon objects, this type of border should be as scalable as anything ever is in Postscript or PDF, and should show up at some high magnification.
I haven't studied what is actually rendered to the XML/HTML workflow, but I'd start by examining the raw mark-up text generated, and looking for border* attributes in the CSS (which might be inline, internal stylesheet, or external stylesheet, and could take some effort to track down).