I'm not sure that I follow your comment about sort order. If you use NVARCHAR, you'll get a Unicode sort on that column (unless you specifically force another collation), which will sort the Arabic text correctly. Did I miss something?
not always sort order is arranging things in the character-code order. say in case of english we have both case-sensitive and case-insensitive sorting & matching (finding). there could be accent-sensitive or accent-insensitive behavior as well. though i do not have ideas about what all and else is applicable for Arabic, but generally i believe it should get mapped with a collation.