TEI Council Meeting 28–29 Apr 05

Times are local time (CEST, +02:00) unless otherwise noted.

Members present were:

  • Syd Bauman
  • Alejandro Bia
  • Lou Burnard
  • James Cummings
  • Matthew Driscoll
  • Julia Flanders
  • Laurent Romary
  • Susan Schreibman
  • Natasha Smith
  • Edward Vanhoutte
  • John Walsh
  • Perry Willett
  • Christian Wittern

Which is to say that all Council members were present (including the ex-officio editors), except for Sebastian Rahtz, who had indicated before the meeting had even been arranged he would not be available.

Meetings were held at AFNOR (room 34), from 14:15 to 18:00 on Thu 28 and from 09:05 to 16:51 on Fri 29, with a lunch break from 12:30 to 13:45. Because her purse had been stolen in Paris, Susan Schreibman was unable to attend on Friday; Laurent Romary left at 12:28 on Friday.

 

Council extends its heartfelt gratitude to AFNOR for hosting our meeting, and to Laurent Romary for organizing it. An essentially flawless operation at a nice, clean, well-run institution helped provide for a successful meeting.

Council also extends its condolences to Susan Schreibman, who had to suffer through the (all too common) frustration, stress, and annoyance of having her passport, money, and credit cards stolen when her purse was snatched in Paris.

With consultation of the group, Chair inserted two new agenda items and swapped the order of two others. The added items were a discussion of equiv, which never occurred, and of edw89. Two items were swapped such that chapter reviews were discussed prior to new work items, in order to accommodate Laurent Romary’s departure.

Review of ActionsThe following action items from the previous conference call were reviewed.

    • Julia Flanders reports TEI now has a wiki, set up by Sebastian Rahtz, Matthew Zimmerman, James Cummings, Daniel Pitti, and others at Oxford. Anyone with valid e-mail address can register and edit the wiki. All that is on the wiki now is the call for stylesheets. By meetings end an area for submission of P5 customizations had been created by James Cummings & Syd Bauman; no customizations were deposited, though. James Cummings and Daniel Pitti will be submitting model stylesheets. There have been two responses to the Call for Stylesheets so far, neither one of which was actually a stylesheet, but one indicated excitement about the plan.

    • Lou Burnard reports P5 HOWTO has been circulated, and that he is seeking feedback on it. John Walsh & Christian Wittern have both read the document & tried the procedure outlined, but each of them had difficulties which have been reported to Lou Burnard.

    • James Cummings reports that the README is only for the p5 module. Christian Wittern suggests, and it was generally agreed, that the README should provide sufficient instruction to facilitate a local installation, which would require the stylesheets module, too. Council discussed whether or not the stylesheet information should be in the same README or a separate one, but decided it was not fruitful to make a decision at this time, as Sebastian is changing how the two modules interact. Create a HOWTO for the stylesheet module however he prefers (as part of current HOWTO or separately) , will be combined or separated later.2005-07-01

    • Julia Flanders reports that she had indeed circulated reports.

    • Perry Willett’s report is on today’s agenda

    • John Walsh reports that the draft report on MARC/TEI crosswalk had been.

    • Matthew Driscoll reports that a set of reviewers for the MS chapter had been contacted; all have responded affirmatively, except for one (a contact at the Ecole des Chartes who would be followed up by Laurent Romary). Consuelo Dutschke has als sent in detailed comments. Council agreed that Matthew Driscoll should review the critiques that he receives, and post as soon as possible a list of items that need to be addressed and perhaps a proposal for further work for Council’s consideration. post summary of comments received on MS, with proposed responses finish revisions to manuscript description chapter (MS) 2005-10-21.

  • Christian Wittern reports that the assignments of chapters to members of Council for review has been accomplished.

The editors both point out that it is very important that all council members become familiar with creating P5 customizations with the ODD language (either using Roma web interface, or otherwise).

Council discussed some details of the documentation needed to make effective use of P5. It was suggested that three different documents are needed:

  • a reference document on the ODD tagset — already done, the DocumentationElements (aka TD) chapter
  • a tutorial guide on using ODD — already begun: edw88
  • a reference document discussing the principles to follow when modifying TEI — the Modifying andCustomizing the TEI DTD (aka MD) chapter, which needs major surgery

Currently instructions for installing and running the standalone Roma utility only apply to a generic (Debian-like) system. It was suggested that this was easily confused with the web-hosted Roma application, and maybe the two should have different names. It was agreed that modifying the standalone shell script to run in other environments would be very useful. attempt to get standalone Roma working locally under Windows2005-07-01

It was suggested and agreed that instructions for using P5 should include some guidance for how to use it in particular environments, for example basic high-level instructions, such as which folder should contain the P5 grammar fragments (DTD or schema); how to tell the application which grammar you want to use for a given document; how to add new grammars etc. Instructions for using a P5 grammar in XMLSpy & XMetal 2005-07-01 Instructions for using a P5 grammar in oXygen 2005-07-01

In response to a question from John Walsh, Council discussed whether users should be advised to specify TEI grammars by defining an ODD or by manipulating grammar fragments in a particular schema language such as RelaxNG or DTD. Both are permitted, but for modifications in particular the consensus was that the ODD route was preferable, assuming the availability of reliable tools.

The subject of extending the manuscript description capability to stone rubbings was raised. Council was reminded that we had explicitly decided to handle manuscripts first, before all other Text Bearing Objects ( TBO text bearing objectss).

Principles of Attribute Usage in P5Council discussed ED W 86 in detail. The first item discussed was the general proposal for handling of textual attributes on elements that have content.

Real text and where to put itA recommendation for an internal wrapper element around the elements that represent what used to be attributes was considered.E.g., something like

<supplied-prime> <attrs> <agent>ablah</agent> <reason>rblah</reason> </attrs> <supplied>whatever</supplied> </supplied-prime>

Use of a generic wrapper element, rather than different wrappers for each content-bearing element which currently has textual attributes (i.e., the equivalent of ISO TC37/SC4’s brack element, was also considered.E.g., something like

<prime> <agent>ablah</agent> <reason>rblah</reason> <supplied>whatever</supplied> </prime>

The editors objected to both of these proposals because it would be impossible to constrain which attributes go with which elements.

It was also suggested that the wrapper element retain the same name as in P4 name, and have a generic content child element. E.g., something like

<supplied> <agent>ablah</agent> <reason>rblah</reason> <content>whatever</content> </supplied>

. This could only work if all elements had the same content: it was agreed that investigation of real cases was needed to decide the issue to create some examples demonstrating the different approaches. This proposal was thought to be sufficiently promising as to merit further investigation. It may also be useful for other similar problems, particularly with the naming set of elements.

generic datatypes and further constraintsCouncil discussed some of the issues surrounding the simplified datatype system suggested in ED W 86.Concern was expressed that relying solely on the use of the valList element limited our ability to constrain attribute values (you cannot define a valList to say that an attribute must be a positive integer less than 42, for example). It was not clear how many such cases there were: if they are few in number, a small set of patterns might suffice for that need.

Overall, Council endorsed the general approach in favour of having fewer, more general, datatypes as presented in ED W 86. review all current attributes, attempt to classify each as a particular datatype, looking for attributes that need a new datatype, are eligible for requiring a <valList>, or need extra constraint; also consider use of tei.tokens vs. rng:text, particularly with respect to rend attribute 2005-06-01.

A suggestion was made that a datatype tei.enumerated be created which would be similar to tei.token, except that it would require a valList (i.e., a documented closed list of possible values), whereas tei.token permits, but does not require, a valList. It was suggested that when Syd goes through the list, he would report if any seem eligible for requiring valList; also if any need extra constraints (like pattern or positive integer).

It was agreed that tei.enumerated might be useful even though a schema designer could just change it to tei.token.

Council briefly discussed whether or not every element that bears a type attribute should automatically get a subtype as well: it was noted that deciding this issue was a matter of how the typed class should be defined.

MiscellaneousIt was noted that some of the elements in ED W 79 are already handled, and thus are not listed in the table at the end of ED W 86 (e.g. the sic attribute of corr).Council briefly discussed, but did not endorse, a suggestion for a capture schema, and a stylesheet for transforming to vanilla TEI.

It was suggested that we look at the three methods so far proposed (and perhaps others) for handling reg on the various naming elements.Those three solutions are:

    • <choice> <name>Bill Clinton</name> <reg>Clinton, William J.</reg> </choice>

      Recommended by ED W 79

    • <reg id=”rn17″>Clinton, William J.</reg> <!– above <reg> element may be somewhere in <teiHeader> or <hyperDiv>, and may be part of a new prosopographic element, and may end up being something other than <reg> … all undetermined as of yet –> <!– … –> <name reg=”#rn17″>Bill Clinton</name>

      Suggested as a possibility by ED W 79

  • <name> <reg>Clinton, William J.</reg> <content>Bill Clinton</content> </name>

look at the various methods for handling reg of the various naming elements, and report back to Council 2005-07-01.

The index element has the three attributes for specifying the index headings under which the current spot should be indexed (level1, level2, and level3). Lou Burnard reported on recent discussion among Lou Burnard, Syd Bauman, Sebastian Rahtz, James Cummings, Christian Wittern, and others on what to do with these attributes; and presented a further suggestion using seg children of label; more to follow. Report on progress of discussions of <index> attributes

Council discussed some of the implications of using tei.tokens rather than rng:text for the datatype of the global rend attribute. It was agreed that this issue should be addressed in Syd’s report on datatypes.

Some open issues

    • ED W 86 asks whether these two perform the same function, and thus whether or not both are needed. Council recommended that certainty of date and exact of dateStruct become precision. Council noted that cert means different things in different places, and thus this issue needs urgent attention. look at cert attribute on different elements and think about how they might be rationalized, and report results 2005-07-01

    • Council suggests renaming key to sortKey when used for that purpose.

    • The ref attribute is declared IDREF when it occurs on some elements, but for others is declared as a token. Council thinks all should be made pointers. Change P5 so that all RESP attributes are declared as pointers. 2005-09-01

  • Canonical references were briefly discussed with respect to loc of app; it seems we could simply remove loc in favor of a cref (which would likely be used in alternation with from & to) but we need to look into location-reference meted etc. research whether or not LOC performs same function as CREF 2005-07-01

It was suggested that examples of usage of various new encodings of the attributes will be useful.

Principles for rationalization of TEI classesCouncil discussed ED W 87.Classes are all declared globally, but are instantiated (i.e., populated) by individual modules. Classes can be hierarchical. Creation of a class system from the bottom up (deciding that a group of elements have a common semantic and thus belong together) and from the top down (finding commonalities in syntax of content models) are both necessary.

Council directly addressed the first three of nine rules for revising the class system.

    • Council changed

possible

    • to

reasonable

    • .

Council agreed, but decided to entertain the suggestion that every attribute (whether 2 exist or not) be in a class.

On suggestion of the editors, Council reworded this to … should reference TEI datatypes (which are themselves derived from W3C datatypes) directly … .

should consider the existing classes, and make suggestions as to what new ones are required, if any2005-07-01

Council to ask Matthew Zimmerman and Wendell Piez if either would be interested in writing a visualization tool to make class system more tractable.

Council decided that a meeting to straighten out the TEI class system should be held; it is tentatively scheduled to occur in Oxford on or about Mon 26 Sep. At least Lou Burnard, Syd Bauman, Sebastian Rahtz, Laurent Romary, Christian Wittern to attend.

Next Council conference call was scheduled for Fri 08 Jul 13:00 UTC.

equivDiscussion of Sebastian Rahtz’s suggestion on use of equiv made in e-mail to Lou Burnard, Syd Bauman, & Laurent Romary on 28 Apr was deferred (but LR has subsequently made some comments).
Associating a Schema with an InstanceSyd Bauman presented ED W 89, with an emphasis on the fact that this is hopefully going to occur in cooperation with at least DocBook, if not OASIS. Concerns were raised that this was not a TEI core concern, since it was an issue common to any XML application. Council agreed Syd Bauman should continue this activity at least until next conference call, when he should report and Council will again decide.
Chapter ReviewsChair expressed concern over the process for activities that include work on a chapter.During chapter review, potential corrigible errors should be reported to the council list.

Reviews may ignore occurrences of global changes that have not been effected yet, including:

  • ID/IDREF
  • DOCTYPE declarations
  • references to additional or base tagsets or the pizza model

However, reviewers should note places where entity references are suggested.

SASyd Bauman reports that the prose in the Linking, Segmentation,and Alignment explaining XPointer and TEI extensions thereof still needs a lot of improvement. finish writing SA discussions of pointing & XPointer extensions; identify external reviewer/s for the chapter. 2005-07-01 Also noted that what used to be a chapter on canonical references (aka CR) is now a section of this chapter; it is not clear whether ot not this should be its permanent placement. Council suggested approaching Michael Beddow as a reviewer for this chapter.
Laurent Romary Laurent Romary reports that the specification of chapter TerminologicalDatabases is done, but prose not written yet. Laurent Romary reports that a member of his research group (Mathieu Quignard) had made some proposals for extension of the chapter on Transcriptions ofSpeech.

Laurent Romary is still working on revising the chapter om PrintDictionaries.

Edward Vanhoutte Edward Vanhoutte suggests that as the chapter Base Tag Set for Prose had been dropped a sentence should be added to a paragraph in Structure of theTEI Document Type Definition describing how the Guidelines can be used for prose. Furthermore, this sentence should be indexed under prose.
John Walsh & Natasha Smith on HD, SH, and CO The editors particularly request that John Walsh & Natasha Smith consider ways of restructuring the material now thrown together in the chapter Elements Available in AllTEI Documents, as this chapter is getting unmanageably large.
James Cummings & PWJames Cummings & Perry Willett report that their current impressions are that:

  • Names andDates may need a lot of work; Council suggested that the Ontologies SIG may be a useful source of input

It was noted that much of the work likely to be needed for ND ties in with task Julia Flanders & Perry Willett are charged with, above. Also classes need attention in this area.

SG does not need to be reviewed at this time, but is still in need of expansion.

Christian Wittern Christian Wittern has deferred his review of the PrintDictionaries chapter, pending Laurent Romary’s work on it. Christian Wittern reports he has made notes on the Default TextStructure and Simple AnalyticMechanisms chapters, and will write them up and post to the council list soon.

Julia Flanders Julia Flanders believes that the chapter Certainty and Responsibility has some areas of underspecification or ambiguity that need attention, and that the chapter provides insufficient capability compared to overhead needed. She plans to do further research into whether the appropriate solution is to abandon the attempt or put the effort into doing it right. Council discussed the fact that this somewhat experimental chapter may in fact be utterly unused. Research into whether or not anyone is using it (or would like to use an improved version) seems worthwhile. Julia Flanders’s initial recommendation for the chapter Base Tag Set forVerse is that Council consider ditching numbered lgs. Most, if not all, members thought this was a good idea. The editors are interested in what else can or should be done in this chapter — should rhymes be encoded more explicitly; are there other useful constructs that should be marked up; more examples are always good.

Alejandro Bia Alejandro Bia reports that the new Languages and Character Sets chapter is excellent. The chapter has lots of typos & mistakes, and thus Alejandro Bia has lots of corrections which he will send to the editors. The language that explains how to use ISO 3066 tags needs improvement; in particular the fact that 2-letter codes are preferred over 3-letter codes when available is not mentioned.
biblItemPerry Willett, Syd Bauman, and Lou Burnard reported that a small group of (2 or 3) individuals have formed xml-biblio, a Sourceforge project expressly to investigate use of TEI for bibliographies. It is not clear whether any members of this group are bibliographers, cataloguers, or have any particular librarian expertise. The group suggested that biblStruct be altered to remove monogr, analytic, and series and to permit self-nesting structures rather than the traditionally flat bibliographic structure They also propose a few extra attributes with which to encode the relationship between nested entries. They propose that the various bibliographic child elements be permitted in any order, but that no PCDATA be allowed. Syd Bauman pointed out that it could no longer be called biblStruct, as it was no longer rigidly structured, and so the new element was named biblItem.The group has developed an examples fileLatest version should be available from . from the

Chicago Manual of Styleexamples. While Syd Bauman has tried to improve it a little, this file still proves to be somewhat unhelpfully inconsistent and poorly encoded.

The main differences between bibl and biblItem are the additions of attributes for semantically informative nesting and the removal of PCDATA. It was pointed out that, were some minor changes to be made to bibl, biblItem would not be needed. It was countered that the exclusion of PCDATA is an important constraint.

There was overall discontent with the idea of having 4 bibliographic citation elements. get back to BIBL using <bibl>, with as little change as possible2005-07-08

teiHeader’s relationship to other standardsJohn Walsh reports that he and Natasha Smith have created a spreadsheet with one entry for each header-element-in-context, and that this the list is being sent to catalogers at IU and UNC so that they can fill in the corresponding MARC & DC fields for each record. Perhaps MODS will be added later. Council suggested that after the catalogers at IU and UNC have had a crack, perhaps others should also be consulted — particularly a European.They are expecting to be able to move on to next phase (i.e., after feedback from the cataloguers) as early as mid-June. write-up a quick report on the progress of the header element mapping to other standards project and post to council list 2005-07-01

physBibAt Council’s suggestion, Julia Flanders did contact Murray McGillivray, asking him to take over as chair of the physical bibliography workgroup; he has expressed interest insofar as he asked to see the PB materials, but hasn’t gotten back to her yet.It was noted that Anne Mette Hansen has finished defending her dissertation, and thus may now have more time to devote to this WG. Julia Flanders is a little bit optimistic that a group may be able to be formed.

Matthew Driscoll reports that Zdenek Uhlir and Jindrich Marek in Prague are working on a page element which contains the description of the page. Jindrich Marek recently presented a paper at a Bulgarian conference on this approach, which he believes is much better than that of msDesc.

Dr Georg Vogeler has recently organized a workshop on the encoding of medieval charters, dealing both with the physical and informational content. He presented a paper on this Charter Encoding Initiative at DRH 2004 (see the abstract), but its current status is unknown.

Council was reminded that the PhysBib group has a clearly defined set of goals: collation & page descriptions. Lots of other nifty things can & perhaps should be done, but not necessarily by this group.

RomaThe new version (installed a few days ago) is broken. The current version is working but there is an incorrect redirect in effect at the UVA server (now fixed). Lou Burnard updated Council on Roma changes.
Stone RubbingsChristian Wittern is interested in seeing TEI support some epigraphical encoding. Julia Flanders recommends speaking to Elli Mylonas, who knows both epiDoc and their rival. Look at current epigraphical standards, contact Elli Mylonas, report on findings to list 2005-07-01 Epigraphical encoding should be considered in light of what MS will be doing with TBOs.
Changes between P4 and P5People are asking about the differences between P4 and P5 on TEI-L. It was pointed out that several documents (P5 publicity page, P5 mission statement, edw81, etc.) need updating.It was suggested that some sort of high-level summary would be appropriate for the present, and that a more detailed comparison, perhaps with some commentary on the advantages of P5 and suggestions on who should upgrade, would not be possible or appropriate until P5 is complete.

WG reportsISO/FS is still awaiting ballot. The chapter will use the xml:id attribute (as opposed to id). There will be a meeting in Warsaw in August.
Council Work ProceduresDiscussion of audibility during conference calls — speak slower and more clearly; headsets & mute buttons are our friends.In order to try to make the calls themselves shorter, we will try reviewing action items via e-mail the day before the call.

Julia Flanders suggests that every report to be considered be accompanied by a question or items for discussion.

Several council members would prefer not to need to pay for conference calls. investigate alternative conference call systems 2005-07-01

Council needs to be responsible for supervising the work being done on the Guidelines.