On 02/26/10 12:08, Bob Cronin wrote:
PMFJI, but your rant caught my attention. Can you point me to further info
on precisely how they differ (as I am in the business of mediating
mainframe-style email exchange between IBM's mainframes and the
intra/Internets and I ought to be up to speed on this issue. We tend to use
the translation tables shipped with the VM TCP/IP product (which themselves
are derived from tables supplied by IBM's Globalization Center of
Competency) rather than the native ones built into PIPE (mostly because we
need to do translations between a lot more codepages/charsets than PIPE
supports). I am wondering if those tables agree with zOS or PIPE.
The offending codepoints:
z/OS OEMVS311 and "iconv -f ISO8859-1 -t IBM-1047"
ASCII EBCDIC
0x0A 0x15
0x85 0x25
CMS "pipe xlate from 819 to 1047"
0x0A 0x25
0x85 0x15
(Somewhat from memory. IIRC.)
The most frequent victims are, e.g. programmers from Linux
platforms who painstakingly hand-code translate tables from
the code page specifications, then find they don't work for
z/OS data because they overlooked a footnote that explains
the deviation.
<RANT>
CMS Pipelines implemented ISO8859-1 <--> IBM-1047 conversion
"by the book". z/OS added a tweak to accommodate the historic
behavior of C compilers, 3215 printers, and CP command separators,
so CMS and z/OS differ. Dammit, if they differ in even one
[pair of] codepoints, z/OS should have defined a different
code page, not added a footnote to the doc.
</RANT>
-- gil
