Re: Empty IA5String

[Mark Smith <mcs@pearlcrescent.com>]

Kurt D. Zeilenga wrote:
> It's my opinion that IA5 String values should have
> non-zero length.  Non-zero length values were
> explicitly precluded for many attributes of
> IA5String syntax (see RFC 1274).  Removing this
> restriction would cause breaking.

And it may break some LDAP implementations, especially some client 
applications, if zero length strings are disallowed.  But that is just a 
guess; I don't think we really know.


> It is debatable whether RFC 1274 implied a general
> restriction that IA5 String values be non-empty,
> however it seems most agree that it not clear what
> the semantics an empty value would have.

I agree there is no consensus on what semantics an empty value would 
have.  We would need consensus on matching (and I do not think that 
would be very difficult to achieve).


> Those in favor of allowing empty IA5 values have
> primarily argued that the empty value is useful
> to represent no values.  Traditionally, in X.500,
> no value is represented by having no values.
> Some have also that an application could regard
> empty values as being a "default" value.
> Traditionally, " " has been used for such purposes.

I can't say I have seen " " used much (but it seems like a reasonable 
choice if zero-length strings are not allowed).


> In either case, it likely that many clients would
> not regard an empty string as having one of these
> semantics, but as simply being invalid.  That is,
> an zero-length 'foo' attribute is far more likely
> to be regards as an invalid foo than as either no foo
> or a default foo or some other foo.

Agreed.  The real question to me is "Do we have clients that rely on 
using zero length strings?"  Because if we, and if the older 
specifications allowed them, them I see no reason to break them just 
because we have now introduced the concept of string prep.  Clearly some 
implementations will have trouble with zero-length strings, but that 
could be fixed over time.


> I also note that a number of programming languages
> do not handle well zero-length values, often be confused
> as no value.

Really?  I think it is possible in any decent programming language to 
distinguish between a zero length string and no string at all.  Maybe I 
only use good languages or maybe I am forgetting something.


> For these reasons, I suggest that the IA5 string syntax
> be revised to preclude zero-length values.

I think we need more data before reaching a conclusion.  Hopefully this 
thread will generate it.

-Mark