Handling bidirectional text (Arabic/English) using iconv

Hi. Wondering if anyone out there can help.

I am trying to convert some bidirectional text (a combination of Arabic and English) to UTF16 and/or UTF8 using iconv and am getting unsatisfactory results. The input data is encoded in EBCDIC 420. The Unicode output is correctly converted, in terms of character-level conversions, but it seems to me that the entire set of characters (both Arabic and English) is reversed.

It's hard to describe what I mean, but it's as if:

EBCDIC 420: The TV station "areezajlA"

becomes:

Unicode: "Aljazeera" noitats VT ehT

which is correct so far as the Arabic is concerned, but not the English.

Now, I believe this means that iconv on IBM i is not implementing the Unicode bidirectional algorithm http://unicode.org/reports/tr9/ which (as I understand it) should allow iconv to recognize which parts of the text are RTL and which LTR and handle them accordingly.

A bit of Googling led me to this: http://publib.boulder.ibm.com/infocenter/zos/v1r11/index.jsp?topic=/com.ibm.zos.r11.bpxbd00/iconvop.htm

which suggests that iConv on z/OS supports some environment variables which can tell it to implement the bidirectional algorithm (I think).

So far, I can find nothing to indicate that these are supported on IBM i. Anyone know?

Thanks