Re: [unrev-II] Architectural Snag - Validation

From: Joe D Willliams (
Date: Mon Nov 06 2000 - 06:34:18 PST

  • Next message: Jack Park: "[unrev-II] Piper and web collaboratories"

    ----- Original Message -----
    From: "Paul Fernhout" <>
    To: <>
    Sent: October 30, 2000 6:02 AM
    Subject: Re: [unrev-II] Architectural Snag - Validation>
    > -Paul Fernhout
    > Kurtz-Fernhout Software

    Joe> > The best idea is to validate a single file or the entire structure in
    > > real time or just whenever you want or need to against an appropriate
    > > schema using an available open validating parser.

    > -
    Paul> This is an interesting observation. You are saying that the techniques
    > for XML validation could be applied to other things with a hierarchical
    > format. (In Kent's ROSE system defined in data and reality this is
    > linked to "executables"). So, one could in effect say that "schemas"
    > apply to some part of an information network to enforce a certain
    > format.

    Joe> I think this is an excellent conclusion. Potentially it is more that
    just format in that the structure, atrtibutes, and content are validated.

    > Difficulties (not necessarily objections) I have with implementing this
    > approach are:
    > * XML is mainly designed (and planned) for hierarchical data. While you
    > can do non-hierachical data (webs) by various abstractions, it seems to
    > me from a limited understanding that XML techniques (XSLT) isn't
    > necessarily designed to do this better than regular coding. (Anyone,
    > feel free to point to good examples of XML use for webs...)

    Joe> When you link 2 or more sets of hierarchal data, then you have a web.
    These sets of hierarchial data can talk to each other.
    XSL provides an algorithmic transformation from one encoding to another.

    > * DTDs and Schemas are useful, but I believe they may not express all
    > the possible ways of organizing even hierarchical data.

    Well, right now we don't need all forms, just forms for nodes having content
    and attributes. Technically, I wonder if this is true. I guess I am sure
    there are corner cases, but I don't think dificulties start until the graph
    becomes acyclic.

    >I think there
    > are (with DTDs at least) some specific issues with artificial limits of
    > what type of repetitive groups one can have in what area of an XML
    > document.

     Joe> Please tell me when you find this limit.

    > Contrast this with for example encoding XML data as Lisp
    > S-expressions and instead of DTDs or Schemas providing Lisp programs to
    > do validation on the expressions. That would be a fully general
    > solution. However, a Schema is easier perhaps for a general editor to
    > interact with for giving proper editing advice than an ad-hoc Lisp
    > program. (Of course, the Lisp programs might be written mainly in a
    > standard form -- approaching something like the XML Schema format but
    > with extensions.)

    Joe> The idea is that the schema provides a way to validate the candidate
    stucture and content before delivering it for use by the OHS (rendering and
    interactivity). Usually this validation is limited to heirarchy, node types,
    and usual relatively simple data types, For more complex, like
    multidimensional data structures, an experiment is under way to use
    ECMAScript regular expressions to validate XML attribute values and content.

    > * (Deeper) XML (and for example RDF) makes it difficult (or impossible)
    > to reference links within a documents. (Yes, there are XPointers, etc.
    > but I am not yet convinced of their generality for what I mean).

    Joe> Is there some kind of picture of referencing links within documents?

    > By
    > contrast, by design, everything in the Pointrel Data Repository System
    > is intended to be linkable. That is -- any relation itself can be linked
    > to and in effect commented on. This is a completely extensible system.

    Joe> Some difference between having a link based repository and a search
    based repository are beginning to make some impression on me, but, I still
    don't see an important point - Why not both. Any information item can be
    identified by unique ID, shared attribute string, links as attributes, and
    links in the content. Attributes and content can be searched using DOM or AI
    tools, and embedded links can be traversed automatically to construct a view
    of content or by the user as the content is browsed.

    > With an XML Schema or DTD it would not be clear how these commentary
    > links would be effected or limited. One might be able to work around
    > this by enforcing a DTD or Schema only on certain links which were
    > linked into a specific validation space (by other links). So, your
    > suggestion is inspiring some more thought on this...

    Joe>validation space = namespace ?

    > In any event, I like your suggestion. Just trying to think through what
    > it means or how to make it implementable. I've been thinking some myself
    > lately on how to use some XML concepts in relation to the Pointrel
    > system (not validation though -- just mainly import/export), based on

    Joe> I would still say, the question is, how to apply to advantages of
    this Pointrel system to information items like those likely to be
    involved in the OHS project.
    > issues of technology and familiarity. If there was an XML interface to
    > some parts of the Pointrel system then it would provide a well
    > understood API for many common operations and an on-ramp for people who
    > understood that technology (as you outline). Your suggestion paints a
    > bigger picture for meaningful integration.
    > I'm always torn on this issue. On the one hand, I don't want to impose
    > any specific "eternal" validation limits to a set of relations.

    Joe> Is this am important point? You really don't know what realtions you
    want to apply to any specific item, you just know that you want to be able
    to find it.

    >On the
    > other hand, unless the relations are structured in some way relative to
    > code intended to search or otherwise process them, the information will
    > be difficult or impossible to use.

    Joe> The relations are structured first by the people who enter items into
    the repository, then by other users, and perhaps intelligent agents.

    Hey Paul, all this is great stuff. From another message I saw that you
    are going into XML and Java. I hope I am right in that the emerging combo
    of XML/DOM/Schema/Java/DataBinding is what I must call a hypersynergism
    that, along with improved 3D visualization will change a lot. Plus, these
    will turn out to be the best choices for Doug's work.

    Best Regards,

    > =========================================================

    -------------------------- eGroups Sponsor -------------------------~-~>
    Create your business web site your way now at
    It's the fast, easy way to get online, to promote your business,
    and to sell your products and services. Try now.

    Community email addresses:
      Post message:
      List owner:

    Shortcut URL to this page:

    This archive was generated by hypermail 2b29 : Mon Nov 06 2000 - 06:47:16 PST