
Steve Thomas <stephen.thomas@adelaide.edu.au> writes:
ALl you need is this:
<div id="ch1" type="chapter" n="1"> <head>CONSULTATION OF DEVILS, AND BIRTH OF MERLIN.</head>
The "rendering agent" can then, if desired, use the type and n attributes to generate the additional "Chapter 1" heading.
Steve
This should work, at least on the vast majority of modern texts. And I agree that the purpose of marking up a text is to markup the content of the text, not duplicate the original layout or typography. PG is producing electronic editions, not electronic facimiles of an original. This is no problem in texts like A Christmas Carol where Chapters are called `staves', but what if the text uses an alternate spelling for the word chapter, or only uses numbers or spells out the number into words? For example. 1 i. Chapter One CHAPTER ONE chapter 1. Chap 1. CH I First Chapter The type attribute should generally use enumerated values so that processing software can understand that all of these different forms of the concept `chapter' are the same. We could normalize all headings, no matter what the original was, but I would prefer to keep the original. In rare cases it reflects authorial intent or is a stylistic element in the overall flow of the work. For A Christmas Carol I would rather use Scott's approach: <div id="ch1" type="chapter" n="1"> <head type="DivLabel">STAVE ONE.</head> <head>MARLEY’S GHOST.</head> Rather than: <div id="ch1" type="stave" n="1"> In this way processing software would understand that a stave is a chapter when it looks up a reference in another work which points to Chapter 1, page 4 in the Carol. I also think that the type of label should be stated clearly. There might be many kinds of labels in a complex document. ------ Ack! Before sending this I had a look through the TEI manual and found this example: <div1 type="book" n="Herod I"> <head>Libro Primo</head> Which is somewhere between Scott's idea and and Steve's. If you used this for the Carol it might look like this: <div id="ch1" type="chapter" n="Stave 1"> <head>MARLEY’S GHOST.</head> This preserves the type value as an enumerated value, but using the `n' value as a text string rather than an integer make's it more difficult for processing agents to understand the structure of the text. I would prefer that the `n' value be an integer and use the head/label approach. I believe that as a general rule attribute values should be used for items which help process a text, or clarify the meaning of a text, rather than for any part of the text which is displayed. The spec defines the datatype for `n' as CDATA. So `Stave 1' is a legal value, but I would seriously consider making the value more restrictive. All in all I think I still like Scott's approach but I'm still open to any better suggestions. BTW: This has been a fantastic discussion and has helped me clarify a lot of details in using TEI which I hadn't completely worked out before. This is very difficult stuff folks, and b/ -- Brad Collins <brad@chenla.org>, Bangkok, Thailand