ZedDist Strawman1
From zedwiki
Contents |
Strawman 1 Overview
- Is based on the Component Pool principle;
- It defines three profiles,
- none of which is text-only (i.e. it uses the total-delegation-to-idpf approach);
- out of which one is the "lightweight audio" profile, contrary to the research committees recommendations;
- out of which one is completely non-extensible, and the other two support varying amounts of extensibility (see Features below)
- It assumes that additional Profiles cannot be added to the standard without spec revision. The idea is that the maintenance committee puts together an "incubation process" where new Profiles can enter the standardization track smoothly outside of the standard itself, but to become formal parts of the standard, a revision must take place.
Short intro to the Backing Framework
Using the Component Pool principle, the three profiles are composed through incorporating the components that the pool requires, and through incorporating some of the optional components, in some cases making them required (as an example, the Audio Profile makes the Audio Channel Component required, e.g it is a instance validity constraint that the presentation contains audio.)
Features
Based on the extension point concept of the Component Pool, and equivalent to the Optional Feature concept in ZedAI, a Profile may activate support for Optional Features. Features can be seen as "plugins" to the core profile model, which typically introduce some new content type (for example, MathML) into the presentation. It should be noted that the activation of a feature does not only change the document model of the Text channel; it can also introduce new constructs into SMIL (the Director) and Navigation (NCX).
In this strawman, it is assumed that two feature extension points will be available in ZedDist:
- The CDR extension point, which allows Features to be designed which are based on injecting foreign material into the text channel using the <object> element and its native fallback pattern. This extension point by design does nothing more than this, in order to be simple to implement (in fact, due to the nature of <object> on the context of a compliant XHTML DOM, the dumbest compliant implementation is to do nothing at the application level).
- The CDI++ extension point (name TBD) which allows composing more major features that impact both the Text Channel, the Director (SMIL) and Navigation. This can also be used to contribute new media channel types? (TBD depends on whether core model of Pro profile includes all channel components by default).
Like Profiles, Features are represented in an online catalog (using the Resource Directory approach as in ZedAI), owned by the maintenance committee. As opposed to Profiles however, new Features can be introduced to the catalog without requiring a spec revision. Generic fallback mechanisms are in place to allow User Agents to fallback gracefully when encountering features they do not recognize, or do not support. (The CDR extension point provides its own inlined fallback mechanism by contract of <object/>; the CDI++ extension point uses container-level fallback)
As can be expected, the package file declares which profile the publication adheres to, and which features (if any) are used within the presentation. Compare [1]
Profile 1: Audio
Nature
An audio book on steroids, allowing page navigation and heading navigation with level support.
This profile has zero extension points activated.
Component usage overview
- Container
- included without any change
- Package
- included without change (package contains identity URIs to declare profile)
- Director
- TBD either use a very simple SMIL grammar a la WMP, or infer order via package spine (or even inject SMIL ns children to package spine). The research committees' recommendation on relying on filename sorting remains, either as a recommendation (because in some locales, filename predictability is legally challenging) or as a requirement.
- Navigation
- Using NCX Modularization, builds one very simple which is exactly one flat list of elements with @targetRole for semantic annotation (see SVN Sandbox). Each navPoint can have
text, audio, and image labels, another reason why this is on steroids.
- Audio Channel
- brings in this component, subsets the allowed codecs to X (and Y?). Makes the inclusion of audio an instance validity constraint.
- Video Channel
- not used
- Text Channel
- not used
- Image Channel
- not used
Profile 2: Classic
Nature
The classic DAISY book with audio, text and image support.
This Profile encompasses the use cases of both DAISY 2.02 and 3.0 content; in the case of audio+ncx, it provides improved navigation, escapability and skippability support as compared to the Audio Profile. It can also be used to represent the "current" Full-text+full audio DTB. It serves as the natural migration path for legacy (2.02 and 3) content.
This profile has the simple CDR extension point activated.
Component usage overview
- Container
- included without any change
- Package
- included without change (package contains identity URIs to declare profile)
- Director
- Declares a SMIL3-based grammar which in terms of feature set is much like the z2005 grammar, however including the "fixes" and enhancements that SMIL 3 brought.
- Navigation
- Using NCX Modularization, builds a default NCX looking much like the z2005 version, but with given enhancements to semantic decoration
- Audio Channel
- brings in this component, subsets the allowed codecs to X (and Y?). Makes the inclusion of audio an instance validity constraint?
- Video Channel
- not used
- Text Channel
- Declares one or two base host grammars that can be used interchangeably per instance (based on outcome of ZedDist_Text_Grammars). Allows the the CDR feature to be used in instances (which basically allows the usage of <object> to refer to external XML Islands, with per-feature stated requirements on inlined fallbacks). Note that referenced islands are not ad-hoc, but must be referring to a defined Feature in the z3986 feature catalog,
- Image Channel
- Included, with a stated set of formats. Instance usage is not a validity constraint.
Profile 3: Pro
Nature
The next-generation, maxed-out DTB. This profile has the simple CDR and CDI++ extension points activated, which means that apart from its relatively rich default feature set, additional major features (such as interactive forms with submission) can be plugged in.
The intent here (in this strawman) is to make all thats needed for video/sign language support part of the core profile's feature set. In other words, forms support would be the first Feature designed for this profile.
(Note that this makes supporting video required for any implementation that claims to support the Pro Profile; an alternate approach is to provide both video/sign language AND forms as Features).
Component usage overview
- Container
- included without any change
- Package
- included without change (package contains identity URIs to declare profile)
- Director
- Declares a SMIL3-based grammar which in terms of feature set is richer than the z2005 grammar (think: switch, video and surrounding logic, y), also including the "fixes" and enhancements that SMIL 3 brought. Activates the SMIL extension point for Features.
- Navigation
- Using NCX Modularization, builds a default NCX looking much like the z2005 version, but with given enhancements to semantic decoration. Activates the NCX extension point for Features.
- Audio Channel
- brings in this component, subsets the allowed codecs to X (and Y?). Does not make the inclusion of audio an instance validity constraint.
- Video Channel
- brings in this component, subsets the allowed codecs to X (and Y?). Does not make the inclusion of video an instance validity constraint.
- Text Channel
- Declares one or two base host grammars that can be used interchangeably per instance (based on outcome of ZedDist_Text_Grammars). Like the classic Profile, allows the the CDR feature to be used in instances. Activates the CDI++ extension point for Features.
- Image Channel
- Included, with a stated set of formats. Instance usage is not a validity constraint.
