Re: Data conversion from IBM to UNIX

In article <43436b$lu_at_news.unicomp.net> brad.stauf_at_chrysalis.org (brad stauf) writes:
>Path:
>noc.tor.hookup.net!hookup!usenet.eel.ufl.edu!news.mathworks.com!tank.news.pipex.
>net!pipex!news.sprintlink.net!news.unicomp.net!usenet
>From: brad.stauf_at_chrysalis.org (brad stauf)
>Newsgroups: comp.databases.oracle
>Subject: Re: Data conversion from IBM to UNIX
>Date: Tue, 12 Sep 1995 13:44:36 GMT
>Organization: greyhound.com
>Lines: 31
>Message-ID: <43436b$lu_at_news.unicomp.net>
>References: <431i9c$iq8_at_aztec.co.za>
>Reply-To: bstauf_at_greyhound.com
>NNTP-Posting-Host: sl11.chrysalis.org
>X-Newsreader: Forte Free Agent 1.0.82

>zedez_at_safmarine.co.za (Ewald Ziervogel) wrote:
>>We are in the process of moving some systems from an IBM/MVS DB2
>>database to Unix/Oracle. We have uncovered a major problem (for us)
>>which I am sure has been solved by others.
>>Some of our data fields are held on the DB2 database as binary, and
>>when transfered to the UNIX platform the conversion from EBCDIC to
>>ASCII screws them up!
>>Also the Oracle load utility will ignore any data fields containing
>>nulls (x'0000' for eg), with disasterous results.
>>Neither Oracle (locally), IBM or HP can cast any light on this
>>problem.
>>Has anyone transfered binary data from IBM to UNIX and if so how can
>>the above be avoided?
>>Any comments most welcome.
>\
>\
>Basically you can't. Do it the manly way ;) and convert the data on the
>maainframe before transferring it. Easy enough in COBOL, just move it from the
>S9 comp sync field to a PIC 9 (no sign or it will flip the last bit and you will
>once again have garbage). Easytrieve or DYL280 probably work the same way but
>I'm not as familiar with them; SYNCSORT and IEBGENER could do it too.

I have been doing MVS to unix/oracle conversions for the past five years and I agree... convert everyting to character on the MVS side and let ftp (or whatever) do the EBCDIC-ASCII character translation.

\lfeges Received on Wed Sep 13 1995 - 00:00:00 CEST