Trials in the Late Roman Republic
Notes on secondary sources

The design question

Two design approaches present themselves:

• bottom up: in the trials, we write references without seeking to link them to a central list of secondary sources (but we DO need to identify them as references). Before publication, we collect all references to secondary sources to create a draft list of references; that draft will need manual editing.

• top down: we create a central list of secondary sources and each reference to such a source in the trials has a pointer to the entry in the list of secondary sources.

The bottom up approach is simpler in the short run, and it's close to the way we have been working, although we have not thus far actually marked references to secondary sources reliably in our trials.

The top down approach is more consistent with the approach used for persons, procedures, courts, and (within limits) ancient sources. It can seem a bit heavy-weight (it does not, for example, seem that linking to person records is easy in the current interfaces), and our experience with ancient sources shows that implementation is not necessarily fast.

Both have in common that they will require that references to secondary sources be identified as such, both in the lists of modern sources (where they are, I think, successfully identified) and in running text like the prose of notes.

On reflection, it seems clear that design consistency should probably outweigh short-term convenience: I cannot find a coherent explanation for a design that creates separate records for persons, procedures, and courts, and hyperlinks to them from trials, but does not do so for primary and secondary sources.

Proposal

1. Create records for both primary and secondary sources analogous to those for persons, procedures and courts. (We do not currently have free-standing records for ancient sources; they should be created.)

2. Hyperlink every reference to primary and secondary sources. (For primary sources, this is existing issue 42.)

3. Distinguish, for each type of entity (trial, person, court, procedure, ancient source, modern source), which of the following applies:

   • immutable: Every reference must take the same form.

   • suffixed: Every reference begins with the same information, but some information (other than a note) may be added at the end. No information is added anywhere except the end.

   • predictable: References may vary, but only in predictable ways (i.e. ways that software can generate automatically given appopriate information). (For example: the first reference to an ancient work will typically display author, work, and locus, unless the preceding source was by the same author, in which case it displays only work and locus; the second reference to any work will display locus only.)

   • variable: References may need to have non-standard forms, for whatever reason. (A common reason for this is a footnote in the middle of the reference, e.g. a footnote regarding a person's gens or praenomen.)

     References of all kinds can be annotated with attached endnotes.

     Trial references are (I believe) immutable.

     Persons are currently treated as variable. It might simplify things if they could be treated as suffixed. (Are offices held in the year of the trial added at the end, or do offices appear in chronological order?)

     Procedures and courts are currently treated as variable, but if we standardize treatment of endnotes it may be possible to make them immutable.

     Ancient sources are predictably varying, at least when they use canonical references.

     Modern sources appear likely to be suffixed, with author and work the common prefix and page number(s) the reference-specific suffix.

4. 4 Adjust the XML representation of all hyperlinks to reflect the classes above. (See existing issue 71 for an earlier form of this task.)

To-do list

• Continue to think about this plan; try to supply explicit mentions of all necessary steps; attach time estimates.

• While thinking, improve the infrastructure to make it easier to search TLRR2 data.

  This is already in our lists as issue 54: we need parallel search interfaces in the public and project-only regions of the site. The public interfaces only have access to (various versions of) TLRR1; the project-internal interfaces need to have access to all databases: TLRR 1, the test database, and the production database.

  Several of the steps outlined below will be simpler if we have project-internal query support on the site.

• Design elements for references to ancient and modern sources in running text. (Perhaps the existing elements will do.) End-notes commenting on the reference should be part of the reference, not simply adjacent to it. This may require retagging some existing source lists.

  For four trials chosen at random (1% of the database), sketch the new markup by hand. Modify the DTD as needed to use this design.

  For forty trials chosen at random (10% of the db), confirm that the design works; modify the design and DTD as needed. (It is desirable but not essential to tag all references in the forty trials; the essential part is to confirm that they CAN be tagged.)

• Study the data to see whether existing references to ancient and modern sources can be identified automatically.

  • If so, write an XSLT stylesheet to do the tagging.

  • If not, write an XForms interface to allow manual tagging of just this one element.

• Using the XSLT stylesheet or the XForms interface just described, tag all existing references to ancient and modern sources.

• Design free-standing records for ancient and modern sources. For each type (ancient, modern), take four sources chosen at random and sketch the new markup by hand. Write DTDs for these new record types.

  Estimate the number of distinct ancient and modern sources.

  For one per cent of each, chosen at random, tag the data by hand; modify the design and DTD as needed.

  For ten per cent of each, chosen at random, confirm that the design works; modify as needed. (Here it is essential to tag all records.)

• Create initial lists of ancient and modern sources and add them to the database. (For ancient sources, lists were created as part of the auto-tagging work; they need to be re-found and modified to match the markup specified. For modern sources, the list of references from TLRR 1 will provide a useful starting point.)

• Hyperlink all existing references to the lists of sources.

  It should be possible to automate this completely or in large part.

• Modify Dexter interfaces to support editing references as redesigned. This will involve:

  • Specifying the ID of the source in question.
  • Adding additional information (for class (b)).
  • Specifying alternate reference form (for class (d)).
  • Changing display stylesheets within Dexter to flag missing source IDs.

• Modify existing display stylesheets to handle new XML elements for references.

• Develop display stylesheets for new ancient-source and modern-source documents.

• Develop XForms interfaces for new ancient-source and modern-source documents.

• Develop list-all queries for sources, analogous to existing queries for all trials, all persons, all procedures, all courts.