TEI vs. Play.dtd

TEI is the leading standard for marking up literature, plays and other non-technical text. (DocBook is the standard for technical works.) However, even the simplified versions may be more complex than is required for most books, especially for volunteer efforts such as Project Gutenberg.

The following comparison is from Martin Mueller's A very gentle introduction to the TEI. I've reformatted it into a side-by-side layout and improved spacing, line breaks, and colors to simplify comparison. The markup on the left is in Martin Mueller's "baby" version of TEI. The markup on the right is from Jon Bosak, one of the inventors of XML. I've included specific comments on each below, plus overall recommendations. (See the original source for Martin's arguments in favor of TEI. I have only excerpted the sample XML.)

Martin Mueller's teixbaby.dtd

Pros

• The header is more detailed (but, as noted below, still missing some useful information).
• div type=act and type=scene: 'n' is an important attribute
• sp (speech): 'who' is an important attribute (but, as noted below, appears redundant here)
• rs key=God and key=StValentine are useful (but, as noted below, the tag name is obscure).
• lg groups all lines of a song (but, as noted below, the information could be attached to sp/speech in the first instance).
• l part=Y and lg part=M may capture useful information (see TEI Lite: Prose, Verse and Drama for details).

Cons

• The header is complicated.
• The header is missing some useful information, e.g. canonical title as a distinct field, credit for the TEI markup, copyright.
• Several fields in the header are "free format" and therefore allow inconsistencies. Additional tags may be useful, e.g. firstName and lastName within author; city and state within pubPlace. (To the extent that the header will usually be generated, the additional tags will not add any complexity for the typical creator.)
• The "text" tag appears unnecessary. (However, DTB has a tag with a similar role: "book"; perhaps in both cases there's a good reason.)
• For castlist: the "div" appears unnecessary; why not put type=castlist on the list?
• For castlist: there's no explicit title. I doubt that "Dramatis Personae" is universal (though perhaps it's the default in TEI Baby?).
• div type=act and type=scene do not indicate that the source uses roman numerals.
• div type=scene: the n=4.5 is confusing. If the "4" is the act, that information is redundant with the previous div and thus a potential source of error. (If the information is useful or required for TEI, it can be generated. It should not appear in the source.)
• div type=scene: no title is included
• stage: hi rend=i for "Enter" and "Song" is presentational markup. That should be avoided unless it's an exception to a general rule. A structural alternative: span class=action.
• sp (speech): 'who' appears redundant with 'speaker' here (if the presentation of the speaker's name is consistent throughout the play, which is probably true for most plays, though rarely true if sp/speech are used in prose)
• speaker: hi rend=i is presentational markup that adds nothing.
• line numbers are useful though could probably be generated automatically
• lg type=song: the "lg" appears unnecessary in the first case; why not put the type on the earlier sp (speech) tag? (That doesn't work in the second case since the song is only part of the speech, but I still think it's a useful simplification.)
• p lb: why are lines specified as line breaks within a paragraph here and one other place, but plain "l" (line) elsewhere? lb strikes me as a poor choice (especially for numbered lines) since it doesn't wrap the line, i.e. it's a tag with no balancing end tag.
• rs key=God and key=StValentine are useful but the tag name is obscure.

Questions

• For castlist: adding id to the list items looks strange. Is it implicitly associating the description with the character?

Suggestions

• Header: get rid of hierarchy that's not useful to PG.
• Header: if the following tags are kept at all, consider spelling out fileDesc, titleStmt, publicationStmt, sourceDesc, bibl. It's more important to have tags that are self-describing than to save a few bytes, especially since these fields are only used in the header.

Jon Bosak's play.dtd

Pros

• The header structure is flatter (but, as noted below, is missing some useful information).
• The document structure is flatter.
• The tag names are simpler and more logical, e.g. "speech" is self-documenting whereas "sp" is not.

Cons

• Uppercase tags have (I think) largely fallen out of favor.
• The header is unstructured, e.g. credit for the markup and copyright.
• The header is missing useful information, at least to the extent this edition can be traced to a specific published work.
• 'FM' is less clear than TEI's 'front' (though I prefer frontMatter to both).
• 'SCNDESCR' is awful. Why not add a description tag to SCENE? If that's not feasible, how about spelling out sceneDescription? It's self-describing and doesn't take up much more space, especially since this tag is not used often.
• 'PLAYSUBT' is awful. If that information is useful (which is not clear here), why not just 'subtitle'?

Questions

• LINE STAGEDIR: is LINE required here? STAGEDIR stood alone elsewhere.

Suggestions

• Add a few tags and attributes to incorporate the advantages of the TEI version without losing the overall simplicity of this version.
• Or, start with XHTML and add tags and attributes from here and (as needed) TEI.
• Dublin Core may be a better model then TEI for enhancing the header.