identity constraints support ?

["Christophe D. Laprun" <ch...@nist.gov>]

Hi all,

Sorry if this has been dealt with before (probably) but archives are
still down. I was wondering what is the status of identity constraints
support. I know it is not supported right now but I'd like to know if
something is going on on this level and if not, why.

Thanks a lot.

Chris

-- 
Christophe Laprun    [Ingenieur ISIMA, France / Guest researcher @NIST]
web: http://www.nist.gov/speech/staff/laprunch.htm
email: chris.laprun@nist.gov
phone: (301) 975 3191             fax: (301) 670 0939
--
The universe seems neither benign nor hostile, merely indifferent -
Sagan

Re: identity constraints support ?

[Andy Clark <an...@apache.org>]

Lawrence Fry wrote:
> Shouldn't the collections/set api in Java 2 be usefull with 
> identity constraints?

Not really.

> If there's a problem of having everything in memory, we need 
> only the set reference to be in memory. Set is an interface 
> which could have an Idimp in xerces. Idimp could be implemented 
> in anyway xerces needs...including not having all the members 
> in memory.

The problem is not the collection of strings that comprise the
identifiers. It's the ability to match the XPath expressions
that *point* to the identifiers in the document instance. In
order to support full XPath for matches, you have to store the
entire document in memory. This is a problem.

And then you also have to have the ability to know what datatype 
the identifier strings are because they have to be compared in 
the value space. So even if you store the entire document, you
have to add info, saving the datatype information as you go.
Otherwise, you just have to do another type of XPath match to
figure out the datatype information so that you can query the
appropriate validator and perform the comparison.

It's a whole can of worms for parser implementors that work in
the real world. In other words, in parser implementations that
don't buffer the entire document. All of these issues and more
will come up again once XML Schema goes to Candidate Rec since
that is the time where implementors give feedback. (IMHO: I
don't think they'll change their minds since the specification
is written in terms of the XML Infoset which assumes that the
document is buffered. A reasonable parser implementation can't 
afford to do this.)

-- 
Andy Clark * IBM, JTC - Silicon Valley * andyc@apache.org

RE: identity constraints support ?

[Lawrence Fry <lw...@bbnow.net>]

Andy,

Shouldn't the collections/set api in Java 2 be usefull with identity
constraints? The spec mentions set:

The value space of ID is the set of all strings.

If there's a problem of having everything in memory, we need only the set
reference to be in memory. Set is an interface which could have an Idimp in
xerces. Idimp could be implemented in anyway xerces needs...including not
having all the members in memory.

Regards,

Lawrence


-----Original Message-----
From: Andy Clark [mailto:andyc@apache.org]
Sent: Monday, August 28, 2000 11:51 AM
To: xerces-j-dev@xml.apache.org
Subject: Re: identity constraints support ?


"Christophe D. Laprun" wrote:
> Sorry if this has been dealt with before (probably) but archives
> are still down. I was wondering what is the status of identity
> constraints support. I know it is not supported right now but I'd
> like to know if something is going on on this level and if not,
> why.

I've been doing a lot of thinking about identity constraints
recently. I have some ideas for making them implementable w/o
requiring the entire document to be buffered but that would
ultimately require a change in the specification by the
working group. I'm all in favor of simplification of XML
Schema, especially in this area.

--
Andy Clark * IBM, JTC - Silicon Valley * andyc@apache.org

---------------------------------------------------------------------
To unsubscribe, e-mail: xerces-j-dev-unsubscribe@xml.apache.org
For additional commands, e-mail: xerces-j-dev-help@xml.apache.org

Re: identity constraints support ?

[Andy Clark <an...@apache.org>]

"Christophe D. Laprun" wrote:
> Sorry if this has been dealt with before (probably) but archives 
> are still down. I was wondering what is the status of identity 
> constraints support. I know it is not supported right now but I'd 
> like to know if something is going on on this level and if not, 
> why.

I've been doing a lot of thinking about identity constraints
recently. I have some ideas for making them implementable w/o
requiring the entire document to be buffered but that would
ultimately require a change in the specification by the 
working group. I'm all in favor of simplification of XML
Schema, especially in this area.

-- 
Andy Clark * IBM, JTC - Silicon Valley * andyc@apache.org