Date:         Wed, 9 Jan 2002 09:55:17 +1100
Reply-To:     "Z39.50 Next-Generation Initiative" <[log in to unmask]>
Sender:       "Z39.50 Next-Generation Initiative" <[log in to unmask]>
From:         Alan Kent <[log in to unmask]>
Subject:      Re: test server SRU
Comments: To: "Z39.50 Next-Generation Initiative" <[log in to unmask]>
In-Reply-To:  <[log in to unmask]>; from
              Larry E. Dixson on Tue, Jan 08, 2002 at 02:36:26PM -0500
Content-Type: text/plain; charset=us-ascii

On Tue, Jan 08, 2002 at 02:36:26PM -0500, Larry E. Dixson wrote:
> Theo,
> I wanted to report that I was able to successfully retrieve and
> display records from your test server.
>
> For example:
>   http://redbelly.bk.nl/cgi-zoek/srw.pl?query=vrml&recordSchema=dc_record&collection=DUTC
> retrieved the following:
>
> <SOAP-ENV:Envelope xmlns:zng="urn:z3950:zng_prototype1" xmlns:SOAP-
>   ENV="http://schemas.xmlsoap.org/soap/envelope/" SOAP-
>   ENV:encodingStyle="http://schemas.xmlsoap.org/soap/encoding/">
>   <SOAP-ENV:Body>
>     <zng:searchRetrieveResponse>
>       <zng:resultSetReference>
>         <zng:resultSetName>200218.108.kqf</zng:resultSetName>
>       </zng:resultSetReference>
>       <zng:totalHits>1</zng:totalHits>
>       <zng:query>vrml</zng:query>
>       <zng:setLimit>3000</zng:setLimit>
>       <zng:records>
>         <zng:record>
>           <zng:recordNumber>0</zng:recordNumber>
>           <zng:recordId>DUTC:4612</zng:recordId>
>           <zng:schema>dc_record</zng:schema>
>           <zng:recordData>
>             <rdf:Description xmlns:dc="http://purl.org/dc/elements/1.1/"
>               xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns">
>               <dc:creator>Margulies, Stephen</dc:creator>
>               <dc:creator>Bauman, Johanna</dc:creator>
>               <dc:title>The power of woe, the power of life : Images of
>                 women in prints from the Renaissance to the
>                 present</dc:title>
>               <dc:date>1999</dc:date>
>             </rdf:Description>
>           </zng:recordData>
>         </zng:record>
>       </zng:records>
>       <zng:status>
>         <zng:statusCode>0</zng:statusCode>
>       </zng:status>
>     </zng:searchRetrieveResponse>
>   </SOAP-ENV:Body>
> </SOAP-ENV:Envelope>

First its great to see people having a go at implementing things. So I
am not trying to criticise - just work out issues as we go along.

But I am a little confused - mainly about goals. I don't have any
immediate plans for SRU (that is, the URL version). But the above
response packet seems to be a mix of SOAP and other things.

For example, I believe with *SOAP* (and SRW) that the most robust
approach for recordData is going to be as a string (I may be wrong).
As such, all the tags would need to be released with &lt; etc inside
the <zng:recordData> element. One reason was that SOAP does not
permit all of XML - for example processing instructions are
not permitted in SOAP, so to allow all valid XML records inside SOAP,
using string is the safest way to go.

The thing with SOAP is that if you use a toolkit, the fact that
SOAP uses XML to encode packets is irrelevant. No-one looks at
SOAP packets. SOAP has lots of weird decoding rules that mean
you would never want to apply an XSLT style sheet to a SOAP packet.
Also, only a subset of the SOAP toolkits support document/literal
encoding (which is the closest thing to the packet above).
So its not an ideal choice in terms of implementations.
So I recommend putting XML in SOAP as a string and let the application
parse the string as XML.

But for SRU it is clearly very useful to return records as XML.
So I think its probably wrong to put SOAP-ENV:Body and SOAP-ENV:Envelope
wrappers around the whole packet.

Maybe what should be done is to define XML schemas for the data
structures, where the schema is reused between WSDL (for SOAP)
and SRU (the URL scheme). But its important to realise that in
SOAP a WSDL schema *does not* define the wire representation of
XML in SOAP packets (confusing, possibly stupid, but true). It
defines an abstract data structure that SOAP then defines its
own encoding rules for (unless you use document/literal encoding).


I have previously put forward the opinion that using strings for
record data is the best approach with SOAP.

However, there is the alternative of document/literal encoding.
One alternative approach that has some attractiveness is to say
- use document/literal encoding for record data
- require that all record data in SRW/SRU have an XML schema defined
- don't allow record data to contain processing instructions

I guess put another way, do people want to be able to insert arbitary
XML documents in as record data to be returned, or is it acceptable
to say with SRW/U that record data must have an XML schema defined
and PI's are not supported.

Oh, there is another ramification by the way - do we want a single
WSDL file (for SRW/SOAP anyway) that represents SRW? If so, then
record XML schemas cannot go in there - because you will have a
different WSDL file per record XML schema. So there will be no
standard file. If you make recordData a string, then you can have
a single standard WSDL file for SRW.

I personally still lean towards a single SRW WSDL file, possible
only by using string for record data (unless someone can show me how!).
In that case, maybe SRU should be based on use of XML schemas and
not WSDL files. I thought SRU was for people without SOAP toolkits
around (just a HTTP client), so you want nice clean XML coming back,
not complex SOAP packets.

Alan