RE: [ubl-dev] infinite loop

["David RR Webber \(XML\)" <david@drrw.info>]

Steve,

Unfortunately that is not what I'm seeing.  Mark is correct vis the NDR for schema.

It is very difficult for tools to automatically detect all recursion, and even when they do detect it - how do they handle it?

For example in NIEM the CAM toolkit detects direct recursion on substitution groups - disables the recursive part - but still includes the non-recursive leaf nodes around and inside the node.  That is very highly specialized handling based on knowing how the human architect engineered those groups.  You cannot expect general purpose schema tools to know all that.

Therefore you are creating the very thing you seek to avoid - inconsistency - because you cannot predict how tools will handle the types of recursion in UBL;  ignore it, fail, or partially include components within the recursion?

When you come back to the business needs - which are for clear and predictable exchange structures - I want to woe you back over to the side of simple here!  That is what I believe we heard today.

We should be recommending best practices technically that lead to a good experience for business users - and not one where they cannot understand why the technology is attempting to push them to use constructs that they cannot fathom the purpose or need for.  Again - I believe association references are clear and the normal practice, whereas recursive components are unexpected.

From the modelling stance - I know this can be tricky in really large models such as NIEM and UBL.  In NIEM I have seen recursion occurring simply because the architects added components without realizing that there existed indirect connections that then meant that part is then recursive.  In NIEM this applies particularly to the "Person" entity - which is now 1.5Mb of XML (wow!) - and includes the life history of a person and their travels through administrative systems managing that!

I would like to see modelling and schema tools explicitly warning architects that the new association they are adding is in fact recursive.  However I see why vendors push back at adding that - because it is non-trivial to put into their products.

Bottom line - if you use associations instead of recursion - you avoid all these pitfalls and issues.  

Thanks, DW

-------- Original Message --------
Subject: Re: [ubl-dev] infinite loop
From: Stephen Green <stephengreenubl@gmail.com>
Date: Fri, June 25, 2010 5:21 pm
To: "Crawford, Mark" <mark.crawford@sap.com>
Cc: "tim.mcgrath@documentengineeringservices.com"
<tim.mcgrath@documentengineeringservices.com>,
"ubl-dev@lists.oasis-open.org" <ubl-dev@lists.oasis-open.org>,
"gkholman@cranesoftwrights.com" <gkholman@cranesoftwrights.com>,
"david@drrw.info" <david@drrw.info>

Such a rule would have to apply to modelers, not software used to generate a schema set.
You would have to forbid, for some reason, a modeler adding, say, any Party association to any Party aggregate.
That seems artificial too.
And out of scope for rules about the schema design. 
And why? Just to cater for the more archaic tools which might not handle recursion?
The tools, I agree with Ken, should these days be able to handle recursion or they are not proper tools for handling XML.
Unusual for me to side with complexity. People have had ample time to get up to scratch with XML by now though.
I sympathise it is complex but technology loves complexity and it adds value, e.g. I would find designing a chip or hard drive hard but I still accept they are well thought out and very useful.
What is more, we all keep buying the software and hardware no matter how complex it gets.

On 25 Jun 2010 21:32, "Crawford, Mark" <mark.crawford@sap.com> wrote:

Tim

There is a specific UBL NDR rule prohibiting infinite loops.

Mark Crawford
Standards Architect
GEPG Standards Management and Strategy
SAP Labs
Mobile: (703) 48...

----- Original Message -----
From: Tim McGrath <tim.mcgrath@documentengineeringservices.com>

To: stephengreenubl@gmail.com <stephengreenubl@gmail.com>
Cc: David RR Webber (XML) <david@drrw.info...

Sent: Fri Jun 25 16:17:40 2010
Subject: Re: [ubl-dev] infinite loop

surely the real practical answer here is that... yes, you can have
infinite loops (as in any hierar...