[xcbf] Re: Biometrics web service

[John Larmouth <j.larmouth@salford.ac.uk>]

Phil Griffin wrote:
> 
> John,
> 
> Looking at a proposed ASN.1 based WSDL web service for
> biometrics for an X9.84 meeting next week in California
> and I have a question. Isn't
> 
>     <xsd:complexType name="PrivacyHTTPResponseType">
>      <xsd:sequence>
>         <xsd:element
>            name="privacyObjectResponse" type="xsd:string"/>
>      </xsd:sequence>
>     </xsd:complexType>
> 
> the same thing as saying something like
> 
>     PrivacyHTTPResponseType ::= SEQUENCE {
>        privacyObjectResponse  UTF8String
>     }
> 
> Seems that it would be. It's just hard to imagine why
> anyone would want to use the former instead of the
> later. Maybe it's what I'm used to seeing, but the
> later would seem to be arguably easier to read, write,
> and comprehend.

Yup.  This is the thrust of most of my ASN.1/XSD comparisons (see my
LineItemPair in the MoU presentation).  The bit I like in the XSD is the
word "complexType"!
 
> Maybe I've overlooked something. I recall your mentioning
> something during a break in Baltimore at the MoU MG meeting
> last week about the ability to carry all of the UTF8
> characters in XML that are allowed in ASN.1 encodings. Do
> you recall this discussion?

This is a very techie issue. XML forbids (if you believe in the letter
of the W3C law) the transmission of control characters except for a very
few such as CR and LF and TAB and SP.  You cannot even (legally) use the
&hex; notation to represent them. This is forbidden.

ASN.1 allows you to write, for example, <bel/>,  or <stx/>, and can
transfer a pure binary string (such as an authenticator) in this way.

**** XSD is incapable of transmitting full UNICODE | ISO 10646, because
XML is incapable of representing control characters (without additional
specification such as the ASN.1 spec). ****
 
Alessandro tells me this has been aired in the XML discussion groups,
and will NOT be mended - they think an XML document should be pure
human-readable graphics characters.

Of course, you can always use a BASE64 encoding for pure binary to
represent a character string, but that is a hell of a lot less efficient
than the ASN.1 transfer with straight Unicode and the <bel/> etc tags
for the occasional control character.

John L