Dane Springmeyer

ugh, the diff is larger than either xml. And many of the differences seem not directly related to the change: http://cl.ly/code/0l1V2S0O0W2n. @tmcw - are you positive this bug is "unavoidable"...

It seems like the next step is to try to create a very small testcase (small stylesheet, only one layer...) and to use it to try to tease out if...

re: condense branch, sweet, happy to try to test re: overall issue of what the heck is ideal/expected - I'm also kind of at a loss. I need to play...

fantastic observation.

Adding more benchmarking statements shows that for the above testcase (https://gist.github.com/859095) most time is taken up in `tree.Definition.prototype.toXML`. It does not appear that any one definition takes a long time,...

okay, back in my court to see if I can gin up a better testcase.

sounds right @tmcw - its basically a list of multiple text symbolizers. Not sure whether you can, on the fly, switch from something like line placement to point. @herm would.

btw, landed in mapnik-reference now and carto 0.9.6 now includes this which will go out with next tilemill (> 0.10.1).

it is a hack and therefore is intentionally undocumented. The only documentation is https://github.com/mapbox/carto/issues/238#issuecomment-19673987. #347 is what it is going to take to make this officially supported.

can you clarify how does this relates to #347?