#443 @resp should be a member of att.global

[Martin Holmes]

This suggestion was originally raised in 2005, and closed because the discussion stalled: http://purl.org/tei/fr/1173968. The idea came up again on the Council list here:

http://lists.village.virginia.edu/pipermail/tei-council/2013/017238.html

I think there are strong arguments for the use of @resp in a wide range of different contexts. As Gaby says, " I can't imagine any element that I would not want to be able to say either who is responsible for the decisions it represents, or from what publication the information so tagged comes."

[Laurent Romary]

Excellent idea. I feel like@source could be given a similar thought and correlated with @resp so that encoders are provided with guidance as to which to use when.

[Martin Holmes]

I keep coming across more people who would like to see this. Another email today: "It's reassuring to know I'm not the only one struggling with... assigning editorial responsibility to a variety of tags".

[Lou Burnard]

The last timne this was discussed (and rejected) I made the point that simply making @resp global won't help with the desire to represent multiple possible structurings. "You can't say the div starts here according to x" at the same time as saying "the div starts here according to Y" simply by using @resp on the <div> -- you need to use <milestone>s or similar, which already have @resp. I haven't heard any new evidence to make me change my mind. We have a mechanism for marking certainty and responsibility in a stand off way which works for all known cases. Why add yet another one which doesn't?

[Martin Holmes]

@Lou: the fact that this proposal won't help with the specific problem you outline (representing multiple possible structurings) is a minor one, surely; it WILL help with a wide range of simple requirements that we face every day. I can't remember ever needing to express the fact that one editor believes a div should start here and another that it should start there, but every day I have to say that X is responsible for this transcribed <seg>, or Y is responsible for this definition.

[Lou Burnard]

Of course, but my point is that we should not make something global
unless we can think of a good use case for its presence globally. If we
can identify some elements for which it would make no sense to use this
attribute (and I would argue that this is the case here) then it should
not be global, and it is up to those who want to extend its use to
specify where it does make sense.

By the way, I am guessing that if you're talking about transcription,
you probably want to add this to att.transcriptional ?

On 30/03/13 16:49, Martin Holmes wrote:

@Lou: the fact that this proposal won't help with the specific
problem you outline (representing multiple possible structurings) is
a minor one, surely; it WILL help with a wide range of simple
requirements that we face every day. I can't remember ever needing to
express the fact that one editor believes a div should start here and
another that it should start there, but every day I have to say that
X is responsible for this transcribed , or Y is responsible for this
definition.

---

[Martin Holmes]

"If we can identify some elements for which it would make no sense to use this
attribute (and I would argue that this is the case here) then it should
not be global..."

I've never heard this argument and I don't think it's legitimate. If we put our minds to it, we can perfectly well think of elements for which we cannot conceive of any need for @xml:id or @xml:lang, to take two examples. When would you need @xml:lang on <pb/>, for instance, or @xml:id on <lb/>? You make something global when the places where it's required cover a sufficiently wide and diverse range of contexts that it's not practical or sensible to capture them in an attribute class.

For the record, in one small project, I need it on <pron>, <seg>, <def>, <phr>, <quote>, <cit>, <persName>, <placeName>, <name> and potentially many other elements.

[Lou Burnard]

The fact that you haven't heard an argument before does not, I hope, ipso facto make it illegitimate! Moreover, as it happens I know of one or two projects which do specify xml:id on <lb/> though I agree that the application of xml:lang to an empty element seems a bit odd.
And while I'm nit picking, saying something is global does imply that it's captured in an attribute class! I think there's a good case to be made for some kind of
att.pervasive class, but stand by my position that we should not hand people a pointed stick such as global @resps that cannot be used meaningfully in all contexts, especially
since we already have a perfectly adequate mechanism in the shape of <respons> ... looking forward to discussing further in providence.

BTW I (and presumably others) cannot see your list of proposed element beneficiaries without doing "view source" in my mailer, so here they are again:

<pron>, <seg>, <def>, <phr>, <quote>, <cit>, <persName>, <placeName>, <name>

[BODARD Gabriel]

In discussion, we felt that a more concrete proposal, with specific use-cases, for what elements or classes of elements need @resp, so that this ticket can be discussed with more focussed and limited proposals.

(It turns out that most of my use-cases are really a desire for the @source attribute, so I'm less help to Martin here than we thought I would be. :-) )

[BODARD Gabriel]

In discussion among TEI Council, it was decided that we should (a) propose expansions to the use of both @resp and @source in tandem, since they are both "talking about attribution of different kinds", and (b) use this ticket to collect concrete and documented use-cases for each (starting with Martin's suggestions of dictionary and naming elements above), with a view to developing a clear proposal for the expansion of these attributes, perhaps in a new class att.attribution.

To get started, several EpiDoc projects have expressed a desire to be able to indicate where various descriptive and historical information about a MS come from, for example, the description of the support is drawn from (but not technically quoted) a particular publication, then <supportDesc> or <support> or perhaps even <objectType> and <material> would be given an @source pointing to the xml:id of the bibl describing that reference. Ditto <history>, <origin>, <provenance>, etc.

[Martin Holmes]

Jennifer Drouin has a case for <castList>, in which "the list of characters in the play is not included in the manuscript given to me by the author, but we thought that the published version of the play should have a list of characters nonetheless (as is, in fact, the case of most of Shakespeare's plays in which the dramatis personae list is an invention of the editor, not found in the original text)."

She also notes that she has had to correct some stage directions, and would like to use @resp on <stage>.

[Martin Holmes]

Expanding on my previous post: in our Nxaʔamxcín dictionary project, we need to ascribe responsibility for:

• Individual pronunciations of an item (<pron>)
• Transcriptions of those pronunciations made in the field (for which we use <seg> inside <pron>)
• Transcriptions which are later narrow versions of phonemic field transcriptions, or phonemic versions of narrow field transcriptions (<seg>)
• Definitions for words and phrases provided by informants in the field and later (<sense>, <def>)
• Example phrases and sentences provided by informants, as well as their definitions, in-the-field transcriptions, and derived transcriptions (<cit>, <quote>, <seg>)
• Division of a form into individual morphemes (<hyph>, <m>)
• Identification of a component of a definition as a suitable short gloss for the entry (<gloss>)
• Assignment of feature structure information to an entry (<fs>)

This may seem excessively pernickety, but this is a project with a 50-year history, and takes place in a cultural context in which appropriate acknowledgement and attribution of contributions, elder authority, etc. are profoundly important.

[Laurent Romary]

What Martin says should be true for any serious dictionary project with more than one editing person. It is very important to maintain a source document that traces back contributions.

[Laura Estill]

I would like to be able to use @resp on \<label>. For instance, I have a seventeenth-century manuscript with some labels from the 17th C and some from the 19th C. Some could also be added by archivists.

[Lou Burnard]

Are these labels or seals or stamps? Anyway, are you describing them as
part of your <msDesc> or are you actually planning to transcribe them
as part of the text?

[Laura Estill]

I'm transcribing them as part of the text (though also mentioning them in the header). They are not seals or stamps.

[Scott Vanderbilt]

I have many, many instances in RIB where various items of metadata, as well as transcriptions, translations, notes, and entries in bibliography and apparatus where @source would be immensely useful in attributing additions and modifications thereto. @resp is not ideal, not only because it is one level of indirection away from the actual authority (usu. a bibliographical item), but also because it is not sufficiently granular when there may be multiple bibl. entries from the same author/editor; @resp alone won't unambiguously point to the actual authoritative reference.

While I have no stand on the global vs. non-global argument, I would very much like to see @source on the following (at the very least): <objectType>, <material>, <dimensions>, <condition>, <layout>, <handnote>, <decoNote>, <provenance>, <location>, <rdg>, and <bibl>.

[Martin Holmes]

Thanks Scott. I would say that wherever there's an argument for @source, the argument can also be applied to @resp, because wherever attribution is being assigned to a bibliographical item, it might perfectly well be assigned to an individual editor who made a decision not based on antecedent authority.

[BODARD Gabriel]

I think I would like to start by, (1) pretty much right now adding @source to att.responsibility, so that it is available everywhere that @resp is, which I believe is pretty uncontroversial (I at least am completely convinced by Martin's arguments to this effect); and then, (2) more luxuriously gradually adding that class to more elements/classes as argued elsewhere in this ticket. The former would have the benefit of making @source available on <rdg> etc., which would be incredibly useful, and it wouldn't prejudice the ongoing discussion of the latter.

Incidentally: why do <respons> and <space> have @resp defined separately, rather than membership of att.responsibility? Is it just so that they don't inherit @cert? We have a better way to do this now, don't we? (I ask because I think I'd want them both to inherit @source as well...)

[Martin Holmes]

@Gaby: re your second question, I see that respons/@resp is only a single pointer, whereas att.responsibility/@resp is multiple pointers. In addition, the definitions are subtly different, although the similarity of the descs suggests this is accidental:

respons/@resp:
(responsible party) identifies the individual or agency responsible for the indicated aspect of the electronic text.

a pointer to one of the identifiers typically but not necessarily declared in the current document header, associated with a person asserted as responsible for some aspect of the text's creation, transcription, editing, or encoding

att.responsibility/@resp:
responsible party) indicates the agency responsible for the intervention or interpretation, for example an editor or transcriber.

A pointer to an element typically, but not necessarily, in the document header that is associated with a person asserted as responsible for some aspect of the text's creation, transcription, editing, or encoding.

I think it would be useful to bring these into alignment anyway, allowing multiple values in respons/@resp, and one obvious way of doing that is to bring respons and space into the class.

[BODARD Gabriel]

Do you think we should break these off into three separate FRs then, to make it easier for us to enact one or two of them even while still debating the third?

1. add <respons> and <space> to att.responsibility (perhaps with @cert deleted, if people care)
2. add @source to att.responsibility
3. make @resp and @source more globally available (i.e. this ticket)

[James Cummings]

I'm of the opinion that we should resist with an iron will making anything that is not truly global att.global.*

I have not been convinced that @resp and @source are truly global.

I agree with suggestion 1 and 2 above by Gabby. I also think there are a large number of elements which may be suitable candidates for membership in att.responsibility.... But I don't want att.responsibility to be global.

-James

[Thomas A. Carlson]

Two use cases have come up for Syriaca.org's Syriac Gazetteer in which @source on <desc> and <note> would have been useful: I wanted to include (1) descriptions of a place taken verbatim from a print source, but a print source which is not in general being encoded in this TEI, and (2) notes about a place taken from a print source, etc. For case (1), I embedded the <desc> element within the <place> element, but since <desc> cannot take a @source attribute, I had to wrap the description text within a <quote> within the <desc>, putting @source on the <quote> element. For case (2), I could have wrapped a <quote> within a <note>, but instead I gave up and used <desc>/<quote> instead, since <note> cannot take a @source attribute.

We would also like to use, parallel to our <desc> element taken from a print source, a <desc> element generated for our project by an editor. This would require adding @resp to <desc> or perhaps to <quote>, but I don't think having <desc>/<quote>/@resp makes sense for our project, since if the description is created de novo it is not quoting some other authority outside of the text.

Thank you all for discussing these issues!

[Martin Holmes]

From Sylvia N. Stoyanova:

"I am encoding the transcription of a manuscript with thousands of internal references, some of which are specific, others less so and some have mistakes by the author. Working with various editors on identifying the most precise destination of links, I need to add the feature of "resp" to attribute proper responsibility for the suggested links, but "resp" cannot be currently contained within a "ref" element or any link describing element that I know of. I also need to add "resp" to marginal notes and other editorial markings (it has worked for dates so far)."