The semantics of OLink are effected by both the OLink tag and the ModeSpec tag.
OLinkOLink is a wrapper around the text which forms the
head of the link. It has four relevant attributes:
TargetDocEntThis ENTITY attribute points to the document which is or contains the link target.
LocalInfoContains the ID of the target in the document specified by TargetDocEnt. LocalInfo is the equivalent of LinkEnd
on the other linking elements.
LinkModePoints to a ModeSpec element which further modifies
the semantics of the link.
TypeIdentifies the link type. The DocBook stylesheets include special processing
for Type=href.
If the link type is “href”, then the HREF attribute from the summary document is used directly, without redirection through a CGI script.
ModeSpecAn OLink element can refer to a ModeSpec element to further modify the semantics of the link. Two aspects
of the ModeSpec element are relevant:
XRefLabelIf the content of OLink is empty, it is possible
to use XRefLabel to control the format
of generated text.
ModeSpec ContentIf the content of the ModeSpec element is not empty,
it is used by the HTML stylesheet in the contruction of the HTML link “href”.
Constructing cross-document links with entities in the authoring system
is all well and good, but how does it work on the web? The heart of the matter
is mapping from the SGML/XML entity in the TargetDocEnt to the base URL on the web, and there are two options: early binding
and late binding.
To use early binding, you must know the mapping from entities to URLs
in advance. The base URL is supplied in the content of the ModeSpec element and that is what the processor uses to construct the HTML
“href”. For example, given the document in Example 1,
the stylesheet would generate something like this for the link:
<A CLASS="OLINK" HREF="http://nwalsh.com/otherdocs/myotherdoc.htm#idval">This is an OLink</a>
Note that the LocalInfo is used
as the fragment identifier.
Example 1. An Early Binding Sample Document
<!DOCTYPE chapter PUBLIC "-//OASIS//DTD DocBook V3.1//EN" [ <!ENTITY otherdoc PUBLIC "-//Norman Walsh//DOCUMENT My Other Document V1.0//EN"> ]> <chapter> <docinfo> <modespec id=otherms>http://nwalsh.com/otherdocs/myotherdoc.htm</modespec> </docinfo> <title>Test Document</title> <para> <olink targetdocent=otherdoc linkmode="otherms" localinfo="idval">This is an OLink</olink>. </para> </chapter>
Late binding delays the resolution of entity to URL mapping until the link is followed. It does this by requiring that a web server process resolve the link. “Out of the box,” the DocBook stylesheets assume that a cgi-bin script on the server called /cgi-bin/olink is responsible for the resolution.
Given the document in Example 2, late binding would produce a link something like this:
<A CLASS="OLINK" HREF="/cgi-bin/olink?pubid=-//Norman%20Walsh
//DOCUMENT%20My%20Other%20Document%20V1.0//EN&fragid=idval">
This is an OLink</a>(without the line break in the middle of the public identifier, of course).
Again, note that the LocalInfo
is used as the fragment identifier.
Example 2. A Sample Document
<!DOCTYPE chapter PUBLIC "-//OASIS//DTD DocBook V3.1//EN" [ <!ENTITY otherdoc PUBLIC "-//Norman Walsh//DOCUMENT My Other Document V1.0//EN"> ]> <chapter> <docinfo> </docinfo> <title>Test Document</title> <para> <olink targetdocent=otherdoc localinfo="idval">This is an OLink</olink>. </para> </chapter>
If a public identifier is used to declare the entity, then it is passed to the link resolution script; otherwise the system identifier is passed.
Even when links span across documents, it's useful to have the stylesheet
generate appropriate cross-reference text. This greatly reduces the “fragility”
of the links at the expense of some complexity in the OLink
processing system.
If you supply content in an OLink, that content is
always used and generated text processing does not apply.
The format of generated text is controlled by the XRefLabel attribute on the ModeSpec pointed to
by the OLink. The text of the XRefLabel attribute is used for the generated text, with the following
substitutions:
Is replaced by the “name” of the link target. This is generally
the “human readable” name of the target element. For example,
if the target is a Figure, %g would
be “Figure”; if the target is a Sect3, %g would be “Section”.
Is replaced by the label (number) of the link target.
Is replaced by the title of the link target.
Suppose, for example, that the link target is the second section in
the first chapter of a book and that it has the title “My Test Title”.
If the XRefLabel attribute contains the
text “see %g %n, %t”, then the generated content
would be “see Section 1.3, My Test Title”.
If the author provides content in the OLink element,
that's the content that is used and no extra processing is required. The sections
that follow each deal with variations in the level of processing required
for generated text in OLinks.
Simple OLinks avoid some of the gory machinery required
to handle general OLinks
at the expense of most semantic variation.
The simple semantics come into play when an OLink
has the following form:
<olink targetdocent="entity"></olink>
In particular, note that there is no LinkMode. The generated text in this case is derived entirely from the public
and system identifiers by the (olink-resource-title)
function. By default, this is simply the title of the document derived from
the description field in the public identifier. The first and last “words”
of the public identifier description field are trimmed off, leaving what is
presumably just the document title. For example, given “-//Norman
Walsh//DOCUMENT My Document Title V1.0//EN”, the derived title
would be “My Document Title”.
The generated text for general OLinks comes from
the XRefLabel attribute on the relevant ModeSpec. The hard part is locating the appropriate replacement
text: the name, label, and title of the element pointed to by the combination
of the TargetDocEnt and LocalInfo.
One way to do this would be to load the TargetDocEnt, find the element with the ID mentioned in LocalInfo and extract the data directly. I chose
something else because I see two significant problems with this approach:
Loading and parsing potentially large documents potentially many times appears to have the potential for significant performance problems.
It would not be possible to form OLink
references to documents written in DTDs other than DocBook.
Instead of loading the actual target document, the stylesheets load a summary of that document's content. For DocBook documents, this summary can be generated by another DSSSL stylesheet, olink.dsl, supplied with the DocBook Stylesheet distribution. The first few lines of the summary for this document is shown in looks like this:Example 3.
Example 3. OLink Summary Document
<!DOCTYPE div PUBLIC "-//Norman Walsh//DTD DocBook OLink Summary V1.0//EN">
<div type="article" name="Article">
<ttl>OLink Semantics in the DocBook DSSSL Stylesheets</ttl>
<div type="sect1" name="Section" id="OLINK-UDERSTANDING" label="1">
<ttl>Understanding OLink</ttl>
<div type="sect2" name="Section" label="1.1">
<ttl>OLink</ttl>
</div>
...The basic organization of the summary document is a nested series of divs and objs with titles (ttls).
Attributes on these elements provide the IDs, labels, and names of the elements.
The GI of the element is also provided. (If you find objects that you think
are missing from the summary, please let me know).
The stylesheets locate this document by resolving the system identifier of the target document and replacing the SGML or XML extension with %olink-outline-ext% (.olink, by default).
If the TargetDocEnt attribute is
missing, it seems reasonable to assume that the link refers to the current
document. This actually offers a feature, you can customize the generated
cross reference text for a particular reference.
When the OLink is an intra-document link, it is possible
to use additional %-substitutions in the cross reference template. See (auto-xref) in dbcommon.dsl for more details.
The examples that follow are all links into the document shown in Figure 1.
Figure 1. Example Target Document
<!DOCTYPE ARTICLE PUBLIC "-//OASIS//DTD DocBook V3.1//EN"> <article id="a1"> <?html-filename olinktarget.html> <artheader> <title>OLink Test Document</title> </artheader> <para>This document demonstrates cross-document linking with OLink. Format this document with the <filename>olink.dsl</filename> stylesheet.</para> <sect1 id="a1s1"> <title>A1 Section One</title> <para>text</para> </sect1> <sect1 id="a1s2" xreflabel="Article Section Two"> <title>A1 <emphasis>Section</emphasis> Two</title> <sect2 id="a1s1s1"> <title>A1 S1 Subsection One<superscript>1</superscript></title> <para>text</para> </sect2> <sect2 id="a1s1s2"> <title>A1 S1 Subsection Two<subscript>2</subscript></title> <para>text</para> </sect2> </sect1> <sect1 id="a1s3"> <title>A1 Section Three</title> <para>text</para> </sect1> </article>
The ModeSpecs in this document are shown in Figure 2.
Figure 2. Example Target Document
<modespec id=ms1 xreflabel="Name: '%g', Label: '%n', Title: '%t'"></modespec> <modespec id=ms2 xreflabel="%g %n, %t"></modespec>
Example 4. An OLink with Content
Source: <olink targetdocent=olinktarget>some text</olink>
Result: some text
Example 5. A Simple OLink
Source: <olink targetdocent=olinktarget></olink>
Result: OLink Test Document
Example 6. An OLink to a Document
Source: <olink targetdocent=olinktarget linkmode=ms1></olink>
Result: Name: 'Article', Label: '', Title: 'OLink Test Document'
Example 7. An OLink to a Section
Source: <olink targetdocent=olinktarget linkmode=ms2 localinfo=a1s2></olink>
Result: Section 2, Article Section Two
Example 8. An Intra-Document OLink
Source: <olink linkmode=ms1 localinfo="olink-understanding"></olink>
Result: Name: 'Section', Label: '', Title: 'Understanding OLink'