Okay, now I see the problem, the genie's already
out of the bottle: "Keep in
mind that <apiname> and <wintitle>
elements are both specializations of <keyword>, so everywhere you
refer to an <apiname> or a <wintitle> it would also be
correct
to just use <keyword>."
And now I see it in the DTD:
<!ATTLIST wintitle %global-atts; class CDATA "+
topic/keyword ui-d/wintitle ">
Why isn't this included in the lang spec definition
(at least, I couldn't find it)? Even in the architecture document it's
only mentioned as an example:
<wintitle class="+ topic/keyword ui-d/wintitle ">A
specialized keyword</wintitle>
Figure 5. Example domain element with class attribute
I read this as an example of something an
implementer might do with DITA - not an specialization inherent in the
base language already.
But now that I understand the situation more fully, I agree your
proposal is better for v1.0.
I can live with ambiguity - especially when it's inevitable.
Thanks for the illumination, Michael.
--Dana
Michael Priestley wrote:
I think that may be a little
stronger
than it can afford to be. Keep in mind that <apiname> and
<wintitle>
elements are both specializations of <keyword>, so everywhere you
refer to an <apiname> or a <wintitle> it would also be
correct
to just use <keyword>. You might mention an <apiname> or
<wintitle>
in discourse without it being a core keyword for the topic as a whole.
I think this gets at what Paul was saying about different explanations
being required when the markup occurs in the <keywords> section
(where
they are keywords for the whole doc) versus in body content (where they
are simply key words of some variety, without being too specific about
exactly why they're important).
How about:
A <keyword> is any word or noun
phrase
that has special significance, for example because it a is a key word
for
the topic, or because it is a part of the user interface or of a
programming
language. Do not use <keyword> when a more appropriate
domain-specific
equivalent is available, for example <wintitle> or
<apiname>.
Specialized elements derived from
<keyword>
may also have extended processing, such as different formatting or
automatic
indexing. If the keyref attribute is used, or some other method of
key-based
lookup based on the value of the element itself, then the keyword can
be
turned into a hyperlink on output (not currently supported).
When DITA topics are output to XHTML,
any <keyword>
elements in the <keywords> element are placed in the Web page
metadata.
I'm avoiding talking about them
being
used for indexing/search when they occur in the body of the text,
because
they may be keywords in the DITA sense without being key words for the
topic as a whole.
Michael Priestley
mpriestl@ca.ibm.com
Okay, maybe Paul's gone home already -
here's
my proposal for a replacement definition for keyword in v1.0 of the
language
spec, my changes/additions in
green:
keyword
A <keyword> is a word that encapsulates the significance of a
topic.
If you searched on a keyword
and retrieved the topic it appears in, you would be happy with the
result.
Being so central to the meaning
of a topic, keywords typically appear inline in the body of the text.
They
can either be marked-up as such
in place, or reproduced in the <keywords> element directly as
metadata.
Compare indexterm, kwd.
Specialized elements derived from <keyword> may also have
extended
processing, such as different
formatting or automatic indexing. If the keyref attribute is used, the
keyword can be turned into a
hyperlink on output (not currently supported).
When DITA topics are output to XHTML, any <keyword> or
<indexterm>
elements in the <keywords>
element are placed in the Web page metadata. In addition, any index
terms
in this context are also used
for supported index processing (for example, for print versions). Any
<keyword> or <indexterm> elements
appearing in the body of the text are also reproduced, uniquely, in the
Web page metadata (not currently
supported).
For ease of reference, here's what's
there
currently (changed section in
red):
keyword
The <keyword> element identifies a keyword or token, such as a
single
value from an enumerated list,
the name of a command or parameter, or a lookup key for a message
(contrast
with term).
Specialized elements derived from <keyword> may also have
extended
processing, such as different
formatting or automatic indexing. If the keyref attribute is used, the
keyword can be turned into a
hyperlink on output (not currently supported).
When DITA topics are output to XHTML, any <keyword> or
<indexterm>
elements in the <keywords>
element are placed in the Web page metadata. In addition, any index
terms
in this context are also used
for supported index processing (for example, for print versions).
--Dana
Dana Spradley wrote:
I agree, Erik - which is why I'd like to
see the <keyword> definition changed in the draft spec to reflect
one single meaning - the meaning implied by the meaning of
<keywords>,
and not a meaning that duplicates the meaning of <kwd> - than two
divergent ones.
After that, I also agree that a userful enhancement down the road would
be replacing <keywords> with something like <topicwords>
or,
say, <semes> or <coverTerms> or
<conceptualBreadcrumbs>
or <searchHooks> or <topicHandles> or something, and
including
all metadata elements that seek to encapsulate the overall gist of the
topic into it, if they aren't put inline in the body of the text
instead.
I'm new to the TC, and hesitate to propose a replacement definition
myself.
Paul Prescod, do you feel like risking a definition in advance of
tomorrow's
meeting? Your offer to try your hand did, after all, begin this
subthread.
--Dana
Erik Hennum wrote:
Hi, Dana, JoAnn, and Rob:
I'd like to submit some reservations about defining elements based on
the
expected output instead of the content semantics:
- An HTML generator could legitimately choose to
populate
the keyword metadata with semantic words that are delimited by
<term>
and <keyword> elements within the text.
- A PDF generator could legitimately choose to
display semantic
words associated with the topic as a whole in the page header.
- A specialization designer should be able to
specialize
an element once to indicate a particular vocabulary (for instance,
<chemicalterm>
or <programword>) regardless of whether an instance of the
vocabulary
appears in the text or is associated with the topic as a whole. That's
possible only if the base element can appear in both contexts. Here's
an
example:
<metadata>
<keywords>
<chemicalterm>molecule</chemicalterm>
<programword>element</programword>
</keywords>
</metadata>
...
<p>You list each <chemicalterm>atom</chemicalterm> in
the <programword>array</programword>....</p>
If anything, I'd suggest that the culprit in creating confusing
expectations
might be the <keywords> element. A metadata <topicwords>
element
that can contain <keyword>, <term>, or <indexterm> might
be better, but I wouldn't expect serious consideration of that thought
until after DITA 1.0
The need for enhancement never ends, and if we try to squeeze in this
one,
I'm sure lots of others will come out of the woodwork (such as the
deferred
<data> element).
Hoping that's useful,
Erik Hennum
ehennum@us.ibm.com
Dana
Spradley <dana.spradley@oracle.com>
erratum: "otherwise you're sending mixed messages..."
In fact, it seems to me this whole discussion was provoked by a bad
definition
for <keyword> in the language spec, which defined it as a keyword
in the technical programming sense, while from the <keywords>
definition
you would have expected it to be defined as in DocBook.
We could solve the entire issue by just revising that <keyword>
definition
to be what <keywords> expects.
If people feel there is a need for keywords in the technical sense to
migrate
beyond the confines of syntax diagrams, then that's a separate issue
for
the folks working on the Programming Domain vis-a-vis the <kwd>
element
- which *is* defined as a keyword in the technical programming sense.
--Dana
Dana Spradley wrote:
I agree on the DocBook part, but
disagree
on the <keyword>
in other
contexts is more like a word from an API or language clause.
The <kwd> element exists for that.
<keyword> outside <keywords> should have the same meaning
it
does within - outwise you're sending mixed messages to authors, and
mixing
up the use.
JoAnn Hackos wrote:
Sounds like a good solution.
JoAnn
JoAnn T. Hackos, PhD
President
Comtech Services, Inc.
710 Kipling Street, Suite 400
Denver, CO 80215
303-232-7586
joann.hackos@comtech-serv.com
http://www.comtech-serv.com
From: Rob Frankland [mailto:robf@rascalsoftware.com]
Sent: Monday, March 14, 2005 9:29 AM
To: 'Paul Prescod'; JoAnn Hackos; 'Don Day'
Cc: dita@lists.oasis-open.org
Subject: RE: [dita] Keywords in DITA
I agree, having followed this thread. Your suggested solution covers
both
use cases. I believe the largest number of users will want the
HTML/Docbook
usage and this enables the programmer writers to meet their needs as
well.
Rob
From: Paul Prescod [mailto:paul.prescod@blastradius.com]
Sent: Tuesday, March 08, 2005 5:46 PM
To: JoAnn Hackos; Don Day
Cc: dita@lists.oasis-open.org
Subject: RE: [dita] Keywords in DITA
Okay, an emerging consensus seems to be that <keyword> in
<keywords>
means <keyword> in the HTML/Docbook sense.
http://www.docbook.org/tdg/en/html/keyword.html
. It is typically hidden from the user as metadata and embedded in the
HTML meta tag.
<keyword> in other contexts is more like a word from an API or
language.
Should we just document it that way? If so, I can suggest some wordings.
Paul Prescod
|