The original Serial Arduino used a polarized capacitor in a particular orientation. It apparently worked fine.
When I designed my alternate PCB, I noticed that the cap was backward, because (I said), it was charged to **-**Vrs232 from the rs232 connector to provide the negative voltage for the rs232 conversion. So I put the cap in the other way, and my board worked too.
When (?) did the S3V3 single sided PCB, he said "WestfW is right, but the cap can also charge to +5V through the transistor when the rs232 cable is not plugged in, so the cap needs to be bipolar." This seems to be the "most correct" evaluation.
But it seems that it's not very critical. Perhaps the voltages involved are low enough compared to the typical cap voltages that normal caps act relatively bipolar anyway. I don't know. If you can't find a bipolar cap, just stick a normal 10uF cap in there any-which-way and see if it works. If it doesn't work, reverse the cap and try again. If it still doesn't work, THEN you can go looking for bipolar caps (although probably the problem is elsewhere.)