Comment 58 for bug 361772

Both filters call Ghostscript to convert PDF into PostScript. The original one (binary) calls Ghostscript with the "pswrite" output device and the new one (script) calls Ghostscript with the "ps2write" output device. The "pswrite" output device turns all text characters into bitmaps whereas the "ps2write" conserves the fonts. The conversion of the characters into bitmaps causes several problems, the squares when the data is fed into the "pxlmono" driver (this bug and the Ghostscript bug #690025 linked above), large output files (bug 44989), and corrupted characters on PostScript printers (bug 362186).

So principally one should prefer "pswrite" over "ps2write", but "ps2write" has also a disadvantage: The output is not DSC-compliant, so page manipulation utilities (like CUPS' pstops filter) will not fully work with these files. In our case it works well, as page m,anipulation was done already before by the pdftopdf filter. After the pdftops filter we only embed Postscript code for the option settings into the job, and this does not need DSC compliance.