Re: [Fwd] perl-ldap on OS/390 ?!

[Chris R. <chr...@me...>]

"R. Reucher" <ren...@cp...> wrote:
> I subscribed to list meanwhile, no need to CC to me...

OK!

> Chris Ridd wrote:
>> One point you might need to consider is that many of the underlying
>> string types used in BER (which is the ISO standard that describes how
>> to encode the LDAP protocol for network transmission) are fundamentally
>> ASCII-based. If you mess with that, you've broken LDAP and won't be able
>> to talk to any LDAP servers :-(
> Correct, and that's part of the problem already ;-(...
>  
>> (Graham, Convert::ASN1's 'make test' should test that strings are being
>> encoded using the correct character sets. Testing "PrintableString" would
>> be a good start.)

When happens when you run this code as is, and when you translate the
'cn=Chris Ridd' :-) string into ASCII:

---8<---
use strict;
use Convert::ASN1;
use Convert::ASN1::Debug;

my $asn = Convert::ASN1->new;
$asn->prepare(q<
		  name PrintableString
	      >);
my $pdu = $asn->encode(name => 'cn=Chris Ridd');

if ($pdu ne "\x13\x0D\x63\x6E\x3D\x43\x68\x72\x69\x73\x20\x52\x69\x64\x64")
{
    print "The value you set for name uses the wrong character set.\n";
}

my $o = $asn->decode($pdu);
Convert::ASN1::asn_hexdump(*STDERR, $pdu);
---8<---

>> So basically everything that you pass into Net::LDAP objects *must* be
>> ASCII or UTF-8 (if you are using LDAPv3.)
> I know that. The question is, _where_ excatly do I need to convert (back
> and forth) ?

I think I said - every value you pass into Net::LDAP from your program and
every value you get out again. Since Net::LDAP doesn't mess around with
values at all, the bytes you pass in should ultimately end up 'on the wire'.

>> I really don't know anything at all about perl on the 390, so this might
>> be a stupid question: what character set is used in the variables in your
>> program?
> It's EBCDIC, which (in more "standardized" form) is "IBM-1047" (as
> opposed to "ISO8859-1", which is ASCII).

ISO 8859-1 is not ASCII. There is an overlap, but ASCII is only a 7-bit
character set and Latin-1 defines 8-bit characters.

But I guess your ISO 8859-1 converter will convert ASCII values correctly,
since ISO 8859-1 is mostly a superset of ASCII.

Is using single-quoted strings different from double-quoted strings?

>> If they're EBCDIC, you *need* to ASCII-ify (or UTF-8-ify) them first.
>> Similarly, you *need* to EBCDIC-ify all values that come back from the
>> LDAP server, as they are either ASCII-based or UTF-8 (if you are using
>> LDAPv3)
> That's what I want to do (or tried to do) using "iconv"... 
> 
> Here's what I did to convert:
> 
> $ascii_data = `echo '$ebcdic_data' | iconv -f IBM-1047 -t ISO8859-1`
> 
> or
> 
> $ebcdic_data = `echo '$ascii_data' | iconv -f ISO8859-1 -t IBM-1047`
> 
> There's also a C API iconv() routine (UNIX98), but that's trickier to
> use in perl... and it should make no difference, basically.
> 
> The problem is, I'm searching for the correct portions of code within
> the Net::LDAP module _where_ I should convert _what_ !?!

From your original message to Graham:
>  1: #!/usr/local/bin/perl
>  2: use Net::LDAP;
>  3: $ldap = Net::LDAP->new('ldapsrv.ourdomain.com') or die "$@";

Add debug => 12 after the hostname to get protocol debugging. This may help
us a lot. Maybe debug => 3 is more useful, as it gives us plain hex.

>  4: $ldap->bind(dn       => 'cn=Directory Manager, o=OurCustomer',

That dn will need to be converted into ASCII.

>  5:             password => 'xxxxxxxx');

That's tricky. The password syntax is OCTET STRING, which means 'just the
bytes, no character set is implied' so whatever the directory server thinks
the bytes are is needed here.

>  6: $mesg = $ldap->search(base   => "o=OurCustomer",

That search base needs changing to ASCII.

>  7:                       filter => "(uid=*)");

That filter needs changing to ASCII.

>  8: $ldap->unbind;
>  9: $mesg->code && die $mesg->error;
> 10: foreach $entry ($mesg->all_entries) { $entry->dump; }

For testing, the dump() method's OK but for real use you'll need to convert
all the values you get back from ASCII into EBCDIC.

> Anyhow, thanks for the answer ! Hope I'll get more input...
> 
> Rene
> -- 
> R. Reucher                     voice:  +49/621/4803-174
> COMPAREX GmbH, VL40            fax:    +49/621/4803-141
> Mannheimerstr. 105             e-mail: ren...@cp...
> D-68535 Edingen-Neckarhausen
> 

I'm guessing that you will want to use umlauts etc in your directory, in
which case I would strongly recommend using LDAPv3 if possible - you need
to explicitly request this in the bind operation - as then all values are
UTF-8.

You should do that because many LDAP servers handle character sets wrongly
(more importantly, differently) when using LDAPv2, so your code will then
not depend on the possibly non-standard behaviour of a particular LDAP
server.

(So whereever I said ASCII above, I mean UTF-8.)

Cheers,

Chris