it might end up being too fragile.

The problem is that GD is really raster-centric, and I am attempting to generate vector output.

I think it would unnecessarily pollute the GD API to bring in vector support.

I'm trying to provide vector-graphics support to people who want to output simple GD-centric programs without burying everyone in excessive dependencies.

Because GD (not GD.pm) works on a pixel-by-pixel basis, I'm trying to work on both vector entities and pixel-related entities at the same time. And GD is turning out to have some nice,convenient capabilities that look handy to keep - otherwise I'd simply rename GD to something else.

I'll see soon enough if this turns out to be a fool's errand, but so far, except for this detail, it's looking fairly good.