RE: I-D ACTION:draft-ietf-ldapbis-protocol-06.txt

["Ramsay, Ron" <Ron.Ramsay@ca.com>]

Steven,

Thanks for the definition.

As you say, audio and jpegPhoto behave correctly, hanks for pointing that
out.

Ron.

-----Original Message-----
From: Steven Legg [mailto:steven.legg@adacel.com.au]
Sent: Monday, 25 February 2002 15:00
To: Ramsay, Ron
Cc: ietf-ldapbis@OpenLDAP.org
Subject: RE: I-D ACTION:draft-ietf-ldapbis-protocol-06.txt



Ron,

Ramsay, Ron wrote:
> I've got to resist this notion of the undefined native LDAP encoding.

I originally introduced the term "native LDAP encoding" as a replacement
for "string encoding" because of the confusion "string" causes.
It means the encoding of values one gets when requesting an attribute
without specifying an explicit transfer option (like ";binary").
The native encoding may or may not be a human-readable character string.

> 
> There are only two encodings for an LDAP value (and I think 
> Kurt was saying
> this also):

Yes, the native encoding and the binary encoding.

It would be more correct to say that the LDAP core specification
defines only two transfer encodings. Others may be defined in the
future.

> 
> o The ASN.1 value inherited from the X.500 definition

The binary encoding.

> o Possibly a string definition defined in RFC 2252 (possibly 
> referencing
> other documents - 2253, 1778?)

... which RFC2251bis refers to as the "native encoding".

> 
> If you request an attribute with ;binary, you get the first 
> one. If you
> request a value which has a string encoding, you get the 
> second one. If it
> doesn't have a string encoding, it is probably one of the 
> certificate types
> and the "fix" is to request it with ;binary.

Saying it doesn't have a string encoding is the same as saying
it doesn't have a native LDAP encoding (unless by "string" you
mean a human-readable representation).

> The types audio 
> and jpegPhoto
> are anomalous (if they are their own string encoding (but 
> then, what is a
> string?), they have no ASN.1 definition, so what effect does 
> ;binary have?).

They do have corresponding ASN.1 definitions, but they're hard to
find since the attributes that use them aren't X.500 standard ones.
I eventually tracked down the ASN.1 definitions by inspecting the
Quipu source code. It turns out both syntaxes are treated as
OCTET STRING for X.500 purposes. The effect of ";binary" on these
syntaxes is the same as the effect on the LDAP Octet String syntax.

> so, the certificate syntax has a native encoding but not a 
> string encoding.

It would appear that you are using "string" to mean human-readable
character string. The Certificate syntax certainly doesn't have a
human-readable character string encoding. However, since LDAPv3
doesn't specify what you get if you ask for userCertificate without
";binary", the native encoding is undefined (or if you prefer, the
Certificate syntax doesn't have a native encoding).

> 
> I think it would be fair enough, in the future, to request it without
> ;binary as any LDAP implementor should realise it is binary. 

It wouldn't be fair enough if in future a human-readable character
string encoding is defined, which is what I've done with the ACI Item
syntax.

Regards,
Steven

> There have been
> no complaints about audio or jpegPhoto. ;binary was simply a BAD idea.
> 
> Ron.
> 
> -----Original Message-----
> From: Steven Legg [mailto:steven.legg@adacel.com.au]
> Sent: Monday, 25 February 2002 12:23
> To: 'David Chadwick'
> Cc: ietf-ldapbis@OpenLDAP.org
> Subject: RE: I-D ACTION:draft-ietf-ldapbis-protocol-06.txt
> 
> 
> 
> 
> David,
> 
> David Chadwick wrote:
> > I think the new text for attribute descriptions is very good. Much
> > better than before, especially the description of 
> subtyping. However,
> > the subsection on Binary Option (4.1.5.1) could still do with some
> > improvement I feel. Firstly, the section is not as clear as
> > recent email
> > we have had on the list, explaining that the attribute 
> value, whatever
> > it is, is treated (or turned into) as an octet string 
> containing BER,
> > and then a TL octet wrapper is placed around it.
> 
> This is not the interpretation Kurt and I have been describing. An
> AttributeValue in LDAP is an OCTET STRING. The contents octets of this
> OCTET STRING are either a series of UTF8 encoded characters (being a
> human-readable native LDAP encoding), a series of arbitrary octets
> (being the native LDAP encoding for, e.g., the Octet String syntax -
> 1.3.6.1.4.1.1466.115.121.1.40) or a BER encoded value (comforming to
> the ASN.1 type corresponding to the LDAP syntax).
> There is no extra OCTET STRING wrapper around the BER encoding.
> 
> By way of example, the contents octets of AttributeValue for the value
> "foo" returned as a value of cn would be (in hexadecimal): 66 6F 6F .
> 
> Returned as a value of cn;binary it would be: 13 03 66 6F 6F,
> the BER for a value of the printableString alternative from the
> DirectoryString{} CHOICE. No OCTET STRING tags to be seen.
> 
> >
> > Secondly, the section talks about overriding the native encoding and
> > sending it as BER instead. But then your example uses
> > certificates, that
> > are native BER encoded.
> 
> The Certificate syntax does not have a native encoding in LDAPv3.
> 
> > So what overriding is there to take place. It
> > would seem like none, and therefore the effect of ;binary is null.
> 
> "Overriding" is perhaps the wrong way to describe it, since for some
> syntaxes there is no native LDAP encoding to override. I 
> prefer to think
> of binary as an *alternative* way of encoding a value. For 
> some syntaxes,
> like Certificate, there is no choice because there is only 
> one alternative
> defined.
> 
> > In
> > other words the section does not clearly indicate to me what the
> > difference would be in transferring a certificate without 
> ;binary and
> > one with i.e. certificate;binary.
> 
> The difference is that "certificate" without ";binary" is illegal.
> The native LDAP encoding is, at present, undefined.
> 
> > This is unfortunate and
> > confusing to
> > the reader. I suggest you use an example that is not a
> > certificate, but
> > is a simple string. You might then also add text to say what
> > the effect
> > of ;binary is on a value that is already BER encoded.
> >
> > I find this sentence particular confusing
> >
> > "Instead the attribute is to be transferred as
> >    a binary value encoded using the Basic Encoding Rules [X.690]."
> >
> > The reason it is confusing is that all values in BER are 
> binary. So if
> > the value is already BER what more needs to be done to it? It
> > would seem
> > nothing.
> 
> RFC 2252 allows a server to store a value in any format it likes.
> If a value is stored as BER then it can be returned in ";binary"
> form without conversion. If stored in the native LDAP encoding
> then to be returned in ";binary" form it must first be re-encoded
> (as BER). The converse applies if the value is requested in native
> LDAP form.
> 
> >
> > Perhaps the sentence might read better as
> >
> > "Instead the attribute value is transferred as an ASN.1
> > OctetStringType,
> > where the embedded octet string value is itself a BER encoded value"
> 
> No, this suggests an extra OCTET STRING wrapper.
> 
> >
> >
> > The second sentence
> > "The
> >    syntax of the binary value is an ASN.1 data type
> > definition, which is
> >    referenced by the "SYNTAX" part of the attribute type 
> definition."
> >
> > also has some errors in it, and could then be changed to
> >
> > "The syntax of the BER encoded value is an ASN.1 data type, which
> > must conform to the "SYNTAX" part of the AttributeTypeDescription."
> >
> > Is that any better?
> 
> I would prefer:
> 
> "The value is BER encoded according to the ASN.1 data type
> corresponding to the attribute's LDAP syntax. The LDAP syntax is
> indicated by the "SYNTAX" part of the AttributeTypeDescription."
> 
> ... with corresponding ASN.1 types nominated for each syntax 
> in RFC 2252bis.
> 
> Regards,
> Steven
>