If we do implement weak constraints for DITA 1.2 or 1.3,
what sort of constraint would we include for task in ditabase (strict or
weak-strict, I assume that general is not an option) and what sort of
constraint would we include in task.dtd (strict or weak-strict)?
Is a weak constraint something that is decided by the
constraint author or is it something that can be overridden in a customized
doctype shell?
Just writing the name weak-strict gave me the creeps, but
I guess we could pick a better name or perhaps task.dtd is always strict and
strict always implies weak, so the name remains task which is strict task or
weak-strict task.
My main concern here is that adding weak constraints is
going to make something that is pretty complicated already even more
complicated. It seems that few people understand it now and so making things more
complicated isn’t likely to help. But I guess if people are willing
to use the doctype shells on “faith” without really understanding
and if the information designers do a good job or are lucky, then people won’t
run into problems and life will be good (as good as it was in DITA 1.0 and 1.1
anyway).
I also worry that making changes like this in a rush at
the last minute without a real proposal may cause us to doing something without
really thinking it through and we may come to regret it later.
-Jeff
From: Michael Priestley
[mailto:mpriestl@ca.ibm.com]
Sent: Friday, September 25, 2009 2:05 PM
To: ekimber
Cc: dita; Ogden, Jeff; rob@ascan.ca; Su-Laine Yeo
Subject: Re: [dita] Why There are Constraints on Conref
Eliot, I think
that's an eloquent description of the rationale for the current situation, and
it's certainly where my thinking is coming from as well.
I did have a
thought on how we might be able to achieve a compromise of sorts, preserving
the strict validation that we both consider necessary for robust processes,
while allowing for broader reuse scope in some of the cases Rob and others have
been asking for (including task->general task).
Here's my
thought:
- if the
doctype developer knows upfront whether a constraint is just an authoring
guideline or whether it's a strict business requirement, they could annotate
the constraint on inclusion
- constraints
noted as "weak" would be ignored by conref validation, and would be
considered unreliable by processors, which would need to assume the constraint
wasn't there as well
Example:
- group A
introduces a constraint that makes title in fig required - they have a figlist
generator that depends on this, so they make the constraint required - if
someone tries to conref from a place that allows title-less figs, they'll get
an error, and will have to negotiate with the other team to introduce the
constraint in their shared content
- group B
introduces a constraint that makes <ul> unavailable in <p>, because
they think it's simpler for their authors. But they don't depend on it, the
output looks the same whether they allow <ul> inside <p> or not,
and they may need to conref content from another team that doesn't use the
constraint. So they make the constraint "weak" instead of required.
Technical
complexity:
- add a flag to
constraints to indicate when they are included as "weak" - eg -
domains=" w(myconstraint) "
- add a check
to conref validation that ignores weak constraints when determining validity
If we wanted
to, we could then implement our strict task as a weak constraint, so it would
be able to conref from loose task.
I think this
might address the concerns of both sides - it gives us strong constraints when
there are business or processing requirements on the constrained structure, but
allows reuse in other cases.
All that said,
I'm very aware of how late in the cycle we are, and I don't think the
cross-constraint use case is actually going to come up that often. So there's
probably a few questions to ask:
- is this a
useful idea?
- does it
address the use cases of both sides?
- if it is, is
it useful enough to change in 1.2, or should it wait till 1.3?
Michael
Priestley, Senior Technical Staff Member (STSM)
Lead IBM DITA Architect
mpriestl@ca.ibm.com
http://dita.xml.org/blog/25
On 9/24/09 10:23 PM, "Rob Hanna"
<rob@ascan.ca> wrote:
>> They might be making content outside
>> their environment unusable *by them*
>> but they would not be making their
>> content unusable to others who use fewer
>> constraints.
>
> What if instead of having fewer constraints, they have different
constraints?
> This is not simply a matter of just two task variations. Over time the
> variations of standard topic types introduced through constraints will
grow.
This is why you have the domains attribute at all: so you can analyze the
constraints at use in two DITA documents to determine if they are
compatible. This at least lets you detect this situation early, rather than
late (such as after a publication has been published).
> For example, consider one possible scenario I see where this might
present a
> problem:
>
> "Over the course of several years a department's style guide
changes several
> times through reorgs and acquisitions - each time enforced with a
different
> set of constraints introduced. There is no budget to convert the older
> content. Over time reuse opportunities are lost and content must be
duplicated
> to suit different doc sets written to adhere to separate
constraints."
>
> As it stands now, constraints do not deprecate gracefully as would
domain
> specializations. Even substructures within structurally specialized
content
> can generally be reused safely.
Note that it is *easy* to remove constraints: you simply update your shells
to stop including the constraint modules. No need to modify the documents
involved.
That means that given two sets of content that have incompatible
constraints
you can make them compatible by changing their governing shells to be less
constrained.
But in fact I think what you are expressing is a problem inherent in any
interoperation situation. DITA isn't changing the fact that two
interchanging communities might have divergent rules over time such that
there may be data and processing interoperation problems: that's an
unavoidable fact of life.
What DITA is saying is "we're giving you a way to detect that case
early,
rather than late" and make an informed decision about how to react.
In practice I think most DITA users will tend to avoid constraints simply
because any constraint, by its nature, tends to impede bidirectional
interchange.
The only reason we're having this discussion at all is because DITA 1.0
inherited an over-constrained task model from IBM, something we should have
fixed in 1.1 but didn't. It's only now that we have the constraint
mechanism
that we can fix the problem (overconstrained tasks) in a way that allows
backward compatibility without having to define a completely different task
base type and specialization.
> This is an issue that can only get more complex over time. In my
opinion,
> processors must be designed to handle the base (or call it general or
loose)
> structures with warnings to ensure seamless exchange of content.
But it's not that easy: there may be processing cases where it would be at
best wrong, at worst impossible, to properly process and render
unconstrained content in a constrained context, because the constraints are
there, in that case, to enable or ensure a particular presentation. So you
can't simply say "processors have to be able to process any
combination of
stuff".
Cheers,
E.
----
Eliot Kimber | Senior Solutions Architect | Really Strategies, Inc.
email: ekimber@reallysi.com <mailto:ekimber@reallysi.com>
office: 610.631.6770 | cell: 512.554.9368
2570 Boulevard of the Generals | Suite 213 | Audubon, PA 19403
www.reallysi.com <http://www.reallysi.com> | http://blog.reallysi.com
<http://blog.reallysi.com> | www.rsuitecms.com <http://www.rsuitecms.com>
---------------------------------------------------------------------
To unsubscribe from this mail list, you must leave the OASIS TC that
generates this mail. Follow this link to all your TCs in OASIS at:
https://www.oasis-open.org/apps/org/workgroup/portal/my_workgroups.php