Re: Empty Directory String - is it legal?

[Jeroen Haak <j.haak@amc.uva.nl>]

Hi,

When trying to import a ldif from an iplanet directory server I
noticed it wasn't possible to have empty values in attributes of the
directory string syntax type.

Since I wanted to know if this behaviour was right, I searched the
mailinglist archive and found the following email:

At Wed, 27 Sep 2000 08:57:06 -0700 Kurt D. Zeilenga wrote:
> At 10:03 PM 9/25/00 -0400, Brent Baccala wrote:
> >Hi -
> >
> >I'm trying to get Microsoft NetMeeting working with openldap, and I've
> >pretty much done it (using the shell backend), but I've still got a few
> >linguring issues.  I need to create a schema and don't have Microsoft's
> >definitions.
> 
> There are some Microsoft schema files in our development tree...
> 
> >Two of the attributes (comment and location) can be left
> >blank, in which case the client will send them to the server with a
> >zero-length value.  I've tried using Directory String for the syntax,
> >but the server barks at the zero-length values.  So...
> >
> >1. should Directory String take zero-length values?
> 
> No.  Directory Strings must not be empty.
> 
> >There's nothing in RFC 2252 to suggest that it shouldn't.
> 
> See X.500.

I couldn't find anything in X.500, so I looked in X.520 for the
attribute definitions.  There I could not find anything that tells the
string must not be empty.  Besides the syntax definition there are
size contrains in some of the attribute definitions.  But that's on an
attribute definition basis, not on a syntax definition basis.

Could you please tell me where in the X.500 I can find that the
Directory String must not be empty?

> >2. should I use Octet String instead?  The app never does searches
> >against this attribute
> 
> The schema specifications I've seen for 'comment' and 'location'
> are syntax directoryString.  I wouldn't muck with these as they
> are likely have far wider use then Net Meeting...
> 
> >3. should there be a server option to turn off the rigid syntax checking
> >and just make these things warnings instead?  
> 
> IMO, no.

It's not wise to do it, but schema checking can be put off.
So why don't make syntax checking configurable?

Regard,
Jeroen.

--
Jeroen Haak, Academical Medical Center, The Netherlands
J.Haak@amc.uva.nl, phone +31 20 5662916, fax .. 5669020