I seem to be unable to return un-corrupted data from a UTF8 Oracle database.

Consider this example:-  (appologies for debugging)

<?
putenv("NLS_LANG=American_America.UTF8");
mb_internal_encoding("UTF-8");
$sql = "select dump(mycolumn) dump, mycolumn from my_table";
$link = OCIplogon("user","pass","connection");
//OCIInternalDebug(1);
$parse = OCIParse($link,$sql);
OCIExecute($parse);
if (OCIFetch($parse))
{
	$mycolumn = OCIResult($parse, "mycolumn");
	$mycolumndump = OCIResult($parse, "DUMP");
	print "found:";
	print bin2hex($mycolumn);
	print ": ".$mycolumn." : len = ".mb_strlen($mycolumn);
	print "<br><br>";
	print $mycolumndump;
	print "<br><br>";
	print mb_detect_encoding($mycolumn, "auto");
	print "<br><br>";
	print mb_internal_encoding();
	print "<br><br>";
	print_r( mb_get_info("all"));

}
?>

e.g.
If mycolumn contains a single UTF8 character E594B4 (three bytes)

I get the result:-

found:bf {upsidedown question mark} len=1

Typ=1 Len=3 228,148,180
EUC-JP
UTF-8

-

I would have expected to see:

found:e594b4 {a japanese glyph} len=3

I suppose the question is, does the OCI interface work correctly with multi-byte strings?