Right-reading, wrong-reading, through-the-glass, chrome-up, chrome-down, digitized-data dark, digitized-data clear, tone reversal, wafer view, reticle view, parity, polarity, ... there's a lot of terminology to keep straight when defining how to build a photomask. But it all comes down to answering these two basic questions:
Most photomask are used mask-side down, or "chrome-down" if the absorber is chrome. While this is obvious for contact lithography, for projection optics the reason is simple enough -- after you use a mask to modulate light into an aerial image, you don't want to send this precious image back across the photomask substrate, losing resolution and intensity along the way. No, you want to feed this directly into your high-NA stepper optics designed to capture the high spatial frequencies diffracting off your mask. For this reason, most photomask patterns are mirrored before they are written onto the mask. The mirroring can be done either during dataprep or on the write tool.
The other key question is what tone to make the polygons in your design data. In theory a maskshop can choose between a positive photoresist and a negative photoresist depending on the answer. In practice the maskshop will standardize on one type of photoresist and reverse the tone of your data as needed. Most photomasks with features of a micron or larger are written with positive photoresist. This is perfect if you want your polygons to be clear but means the maskshop will reverse the tone of your data when you want the polygons to be dark.