The issue is that there are non-topic elements (questions and objectives in this case) that need to be managed and referenced *as though they were topics* even though they cannot be topics.
That is, topics are inherently “contextless” (in that they are, in general, usable in any context and are intended, more or less, to be written that way).
Most non-topic elements have some context, in that they either contribute to the narrative flow of the topic that contains them or are metadata that applies to their container.
It is only L&T questions and objectives that seem to have the quality of being contextless, like topics, but, unlike topics, do not have natural titles.
It is the way that questions tend to be used, for example, to construct many different tests from the same pool of reusable questions, that leads to the desire to use them from maps rather than via conref from within topics.
Chunking does not use topics without their titles—chunking simply allows you to treat topics as though their source file organization was different from what it actually is. It doesn’t change the fundamental processing rules for topics. (Submaps, by contrast, do not have their titles reflected in output in normal processing because the effect of a map reference is as though it was a topicgroup that occurs at the point of reference).
From: David Hollis <email@example.com>
Date: Thursday, November 30, 2017 at 8:36 AM
To: Amber Swope <firstname.lastname@example.org>
Cc: Eliot Kimber <email@example.com>, dita <firstname.lastname@example.org>
Subject: Re: [dita] Managing "contextless elements" as for L&T Assessments and Objectives
Hi Amber, Eliot,
Isn't the generic DITA topic already quite "contextless"?
I thought that chunking was the way to use a topic without the title? I assume there is a title somewhere, just it is not required for the "contextless" topic. This seems like a parallel to Information mapping?
I would be concerned if the proposed solution would be to put content in a map, and would suggest a wider discussion about what the role of a map and a topic should be, going forward. I appreciate that metadata in a map is already used for various purposes: reuse and subject schemes, for instance. But one of the potential solutions would seem to go further than that. I acknowledge Eliot's caveats about this potential solution.
There might be a parallel with resource-only "container topics", or "warehouse topics" that simply provide content for reuse. The topic itself never appears in the final document.
Thanks for the summary, Eliot.
Although a single question is a unit that can be understood in isolation and used in multiple contexts, it doesn't meet the definition of a topic in that it does not natural title. This means that they fall into the gray area that you are calling “contextless elements”. I can't think of a better way to describe it and thank you for your suggested terminology.
Because assessment and learning objectives are fundamental to learning content, we are definitely going to address this issue on the L&T SC. However, we don't want to do this in isolation, so if you have grappled with content that meets the same criteria, please share your experience with the TC.
From: email@example.com [mailto:firstname.lastname@example.org] On Behalf Of Eliot Kimber
Sent: Tuesday, November 28, 2017 8:15 AM
Subject: [dita] Managing "contextless elements" as for L&T Assessments and Objectives
One of the requirements that has emerged in the use of Learning and Training questions (interactions) is the need to manage individual questions as reusable objects through maps.
The typical use case is constructing a test at the map level, rather than using conref.
Using maps for this makes a lot more sense as pretty much all of the available DITA-aware tooling is optimized for map-based reuse rather than conref-based reuse at the sort of scale represented by a test. Imagine, for example, a test consisting of scores or hundreds of questions. Trying to manage that through conref becomes complicated and tedious, at least with typical authoring tools.
This leads to a situation where we want to treat individual questions as though they were topics from a management and aggregation standpoint but the questions themselves are just elements within topics and would not be appropriately modeled as topics themselves (or example, most questions don’t have natural titles).
All of us in the L&T community have struggled with this sort of topic/not-a-topic nature of questions (and to a lesser degree, learning objectives). For example, in the past I’ve defined a “question topic” specialization where the title is an empty element and the navigation title is required. This is a hack but it kind of works but it isn’t really ideal.
In thinking about this more in the context of our latest L&T subcommittee discussions, I realized that L&T assessments are what I would call “contextless elements”, meaning elements that are not topics but that are fundamentally independent of any context.
That is, an individual question is not normally dependent on the elements around it the way that say paragraphs or lists or even figures or tables are.
Or maybe more usefully, questions do not contribute to any narrative flow—they are true objects intended to be combined arbitrarily with other questions.
Learning objectives have the same quality—they do not contribute to any narrative flow and are intended to be used directly or indirectly to characterize different learning content. For example, it is useful to both use objectives by conref and to use more indirect association between objectives and learning objects, such as via reltable or subject scheme.
I haven’t been able to think of any other kind of information that would have this “contextless” characteristic but it seems likely there are such things in other domains, so I think it’s useful to try to capture the abstraction.
So if we accept this concept of “contextess elements” then my main question is what is the best way to enable their direct use via maps?
In the case of creating a test, I want to construct a map where I have topics that represent the test as a whole, for example, to provide initial instructions, introductory material, examples, etc., and also to organize the questions hierarchically (imagine a typical standardized test with many sections and subsections) and then have references to questions where the referenced questions are treated as though they were part of the body content of the parent topic (as defined in the map) for presentation purposes.
One possibility would be to define a new generic topic type of “contextless element container” which has the specific semantic of “this topic exists only to contain content that can be combined together arbitrarily and its title is not to be used in normal output”. That would essentially codify the “question topic” type that I’ve defined in the past. I think this is probably the most appropriate solution because it binds the semantic directly to the topic, rather than to references to it.
Another approach would be to provide some way to say on the topicref “use the shortdesc, abstract, and body of this topic but don’t use the title”, making it clear that the intended presentation result is equivalent to having the same topic content occur in a single topic as though via conref. This is basically what people have been doing to date one way or another, but via private convention and custom processing.
Yet another approach would be an architectural change that allows direct use of elements from maps, maybe using a new “elementref” element or something. I’m not necessarily suggesting this is a good idea, just enumerating the possibilities. This kind of element would raise a whole raft of issues around resolving references from the element to other things based on use context and so might require specific restrictions (for example, such elements must be the abstract or only body child of their containing topics). However, I know there is a latent desire to able to use map-style linking to create the effective content of topics rather than using conref within the topics, so maybe it’s worth considering if only to decide the complexity of the solution is too great.