Federal CIO Council XML Web Services Working Group

1
Summary of Native XML Databases Thread
Joseph M. Chiusano Booz Allen Hamilton
Federal CIO Council XML Web Services Working
Group Washington, DC July 22, 2003
2
The Native XML Databases thread began on 5/12/03

• Brand sent information on Federal Computer Week
  article interview to listserv
• This sparked a thread of 40 e-mails, with main
  sub-threads of
• Native XML Databases
• XML and Semantic Clarity
• XML and Programming Logic
• Each sub-thread is summarized here

3
Native XML Databases
4
A native XML database stores XML in native form

• Native XML databases are employed mostly for
  document-centric applications
• The XML Web Services Working Group currently has
  a pilot project involving native XML databases
• XML Data Exchange Across Multiple Levels of
  Government Using Native XML Databases
• Main players in native XML database arena
  eXcelon and Tamino (commercial) and Xindice,
  eXist, 4Suite, and ozone (Open Source)
• Middleware products that transfer to/from
  relational databases main players JAXB, .NET,
  Delphi, and WebSphere (commercial) and Castor,
  JXQuick, Zeus, and Zope (Open Source)
• Ronald Bourret, XML and Databases, XML 2002
  Conference Tutorial, 12/9/03

5
A hybrid native XML/relational approach is
preferred in some cases

• "Native XML databases add indexing information to
  XML repositories and they can be a good choice if
  you want to retain the original XML
  representation of the information, provided you
  only need limited processing of the information
  therein. Use a hybrid approach (native and
  relational) for XML storage when you need to
  perform indexed operations such as searching and
  aggregation against the information, but also
  need to retain the XML form in which the
  information first entered the system."
• XML Design Handbook, Wrox
  Press, 2003

6
The main disadvantage for native XML databases at
this time is a lack of maturity in mechanisms for
querying/reporting on XML data, compared to
relational mechanisms

• The W3C XQuery Requirements specification is
  still a W3C Working Draft at this time
• http//www.w3.org/TR/xquery
• Additionally, people have become comfortable with
  their RDBMS systems and value the maturity level
  of the major DB vendors
• They may also be hesitant to migrate
  fundamentally relational data (like catalog-type
  info) into native XML stores
• We will continue to see interest in products like
  Oracle that manage both relational and
  hierarchical sets
• Recommended book XML Data Management Native
  XML and XML-Enabled Database Systems, Chaudhri,
  Rashid, and Zicari, Addison Wesley Professional,
  March 2003

7
XML and Semantic Clarity
8
This sub-thread revolved around the notion of
semantic clarity of XML tags in various scenarios

• Example of XML generated directly from a
  database
• ltSTOCKITMgt
• ltNSNgt1234-12-123-1234lt/NSNgt
• lt/STOCKITMgt
• Fundamental question is this good XML?
• What is good XML?
• There were disparate opinions on this question

9
One side represented the developer-oriented view

• Theres no such thing as better XML or worse
  XML, as far as tag names and organization go
  it all depends on the circumstances
• Suppose we are writing XML messages to be passed
  between two proprietary systems, with hundreds of
  thousands of messages being passed back and forth
  each day
• Is this representation
• ltmessagegt
• ltSystemProtocolgt
• ltSystemMessagePartOnegt17lt/SystemMessagePartOne
  gt
• ltSystemMessagePartTwogt33lt/SystemMessagePartTwo
  gt
• ltSystemMessagePartThreegt22lt/SystemMessagePartT
  hreegt
• ltSystemMessagePartFourgt38lt/SystemMessagePartFo
  urgt
• ltSystemMessagePartFivegt41lt/SystemMessagePartFi
  vegt
• lt/SystemProtocolgt
• lt/messagegt

10
One side represented the developer-oriented
view (contd)

• Suppose we are writing XML messages to be passed
  between two proprietary systems, with hundreds of
  thousands of messages being passed back and forth
  each day (contd)
• better than this representation?
• ltmgtltp1gt17lt/p1gtltp2gt33lt/p2gtltp3gt22lt/p3gtltp4gt38lt/p4
  gtltp5gt41lt/p5gtlt/mgt
• In this situation, the most important factor is
  brevity of the message
• If we get too fixated on the human-readability
  aspects of XML, we can lose sight of the higher
  goals of performance, reusability, scalability,
  etc.
• Context matters!!!

11
The other side represented the business
analyst-oriented view

• Human readability and semantic clarity of tag
  names are paramount
• Earlier example would be better represented as
• ltSupplyMaterielItemgt
• ltFederalClassificationCodegt1234lt/gt
• ltNationalIDgt12-123-1234lt/gt
• lt/SupplyMaterielItemgt
• This version applies information analysis and ISO
  11179 rules to produce a more semantically clear
  representation
• This is more inline with the 10 design principles
  of XML
• XML documents should be human-legible and
  reasonably clear
• Terseness in XML markup is of minimal
  importance

12
The design approach used with XML should fit the
task at hand

• Just like you wouldn't use Visual Basic to write
  a device driver, you shouldn't try to use one
  particular design approach to XML to solve every
  problem (no matter what the designers originally
  said)
• The Scalable Vector Graphics (SVG) W3C
  specification is an example of tags that are
  meant to be machine-processable and not
  human-readable
• Examples
• ltrect id"RectElement" x"300" y"100"
  width"300" height"100"
  fill"rgb(255,255,0)"gt
• ltanimate attributeName"x" attributeType"XML"
  begin"0s" dur"9s" fill"freeze"
  from"300" to"0"/gt
• ltset attributeName"visibility"
  attributeType"CSS" to"visible"
  begin"3s" dur"6s" fill"freeze"/gt

13
XML tags cannot by nature convey sufficient
semantic clarity

• What does a tag of LastName mean?
• There is no way for XML to indicate by use of a
  LastName tag that the element is equivalent
  to the concept of a family name
• Order of names is culture-specific
• The only solution is to explicitly document the
  meanings of elements and attributes
• There are various approaches to this, such as
• XML Schema documentation
• Data Dictionaries
• Semantic Registries
• Metadata Registries

14
Keeping tags human readable might be more
expensive initially, but

• It can lead to less maintenance downstream
• It can lead to greater reuse
• It can lead to cost savings

15
XML and Programming Logic
16
There are certain specifications that use XML in
a programming sense e.g. to represent methods

• Example from OASIS Directory Services Markup
  Language (DSML) Version 2
• ltsearchRequest dnOUMarketing,DCExample,DCCOM
  /gt
• This appears to move XML toward being a full
  object model rather than exclusively a
  serialization
• There are also cases in which schemata represent
  IF/THEN constructs as XML
• Is this a good idea?
• If so, there should be some common approach to
  this use of XML

17
This is a wave that began with XSLT and is
becoming more prevalent today

• More people are seeing the value of a
  "declarative programming" approach in which much
  of the processing is "pushed down" into
  processing engines rather than being stipulated
  command-by-command as with 3GL languages
• Examples
• OASIS Business Process Execution Language For Web
  Services (BPEL4WS)
• Web Services Choreography Interface (WSCI)
• W3C XForms

18
Questions?