Metadata consistency: 23 DCLP records, 8 corrections

[tonclap]

This is a batch of metadata corrections to DCLP, found by checking records
against the rest of the base rather than against papyrological judgement:
each one is a record contradicting itself or the identifier it carries.
The 23 records fall into 8 groups, one commit each, so they can be
reviewed — or rejected — separately.

Everything here is internal evidence. No new attribution is proposed, no
reading of any text is changed, and no date is asserted that the base does
not already assert elsewhere.

1. Aristophanes' Clouds filed under Philo Judaeus (7 records)

TM 59256, 59261, 59262, 59269, 59275, 59287, 65547 name Aristophanes com.
in <author> and point @ref at urn:cts:greekLit:tlg0018, which is Philo
Judaeus. Aristophanes is tlg0019, which the other 57 Aristophanes com.
records carry.

These seven are not seven unrelated slips: all seven are papyri of the
Clouds (six titled Nubes with line ranges, one "Opening Lines of
Clouds"). An eighth witness of the same comedy, TM 901286, is filed
correctly under tlg0019.

In TM 59262 and 65547 the same @ref also lists cwkb.org/author/id/765.
That identifier is Aristophanes — it co-occurs with tlg0019 in 19 records
— so those two records already pointed at the right author with their
second identifier.

2. Ulpian filed under Justinian (7 records)

TM 62939, 62940, 62941, 62944, 62945, 63036, 64840 name the Latin jurist
Ulpian in <author> and point @ref at urn:cts:greekLit:tlg2734, which
is Flavius Justinianus — roughly three centuries later.

This is not a house convention for Latin legal texts: across all records
with Latin in <langUsage>, other Latin jurists carry their own
identifiers (Aemilius Papinianus = tlg2374, Iulianus antecessor =
tlg4053), and no other jurist sits under tlg2734 — the 16 records that
belong there are Justinian's own. Within TM 64840 itself, Paulus iuridicus and Pomponius are cited in the same passage with no @ref at
all.

The ref is removed, not replaced. tlg2604 is a different Ulpian (the
rhetorician, commentator on Demosthenes); the jurist may have no greekLit
identifier at all. An <author> with no @ref is not a degraded record
here but ordinary practice: of the 9800 DCLP records that carry an
<author> at all, 4376 — nearly half — carry no @ref on any of them.
It seemed better to leave the identifier absent than to assert a second
wrong one.

A related record is deliberately not included: TM 65841 also carries
tlg2734 on an early record, but there the text reached us through the
Digesta and the overview says so ("letter of Severus Alexander, later
included in Digesta 049 1 25"). That is a convention, not an error — it is
what tells the two cases apart.

3. P.Herc. 573 filed under the tactician Polyaenus (1 record)

TM 130194 carries <title>Adversus Aristonem</title>, a work of Polyaenus
of Lampsacus, the Epicurean, which fits both the Herculaneum library and
the record's own date (200–101 BC). urn:cts:greekLit:tlg0616 is Polyaenus
Macedo, the tactician of the 2nd c. AD. The same base holds TM 59495,
Demetrius Lacon's polemic against that same Epicurean.

Removed rather than replaced, for the same reason as above: the Epicurean
has no identifier in the open corpora. The date is not touched.

4. P.Köln 7 305: Hellenica filed under Xenophon's letters (1 record)

TM 63009 carries <title ref="trismegistos.org/authorwork/1052">Hellenica</title>
with @ref at tlg1754, the separate corpus of Xenophontis Epistulae.
Corrected to tlg0032.

5. A dropped minus sign in three date ranges (3 records)

TM 154374, 243958, 382584 carry notBefore later than notAfter. Unlike
the swapped ranges of #649 these are a lost minus, and swapping them would
be wrong:

• TM 154374 labels itself "99 - 1 v. Chr." — both bounds are BC;

• TM 243958 (P.Herc. 671) and TM 382584 (P.Herc. 153) are Herculaneum
  rolls, which cannot postdate the eruption of AD 79, as a swap would make
  them. For TM 382584 the resulting -0199..0079 is exactly what TM
  942310, 942311 and 942312 carry, spelled out there as "199 v.Chr. - 79";
  for TM 243958 the resulting -0199..0050 matches the five neighbouring
  Herculaneum records TM 243960–243964, which carry -0200..0050.

  (Those five label that range "200 n. Chr. - 50", i.e. AD where the
  attributes say BC. That is a separate inconsistency, in the label rather
  than in the range, and this batch does not touch it.)

6. One genuinely inverted range (1 record)

TM 67691 (Augustine, Sermones, a Carolingian manuscript of the 8th c.):
notBefore="0775", notAfter="0725". Same class as #649, and as there the
human-readable label is updated together with the attributes.

7. Two records declared Latin are written in Greek (2 records)

TM 62721 (P.Oxy. 53 3701) and TM 62722 declare only <language ident="la">
while their editions are Greek throughout and contain no Latin letters at
all: the only Latin characters anywhere in <div type="edition"> are the
word "Column" in the <head> of each column. la replaced with grc.

8. A Graeco-Latin glossary declared Greek only (1 record)

TM 171906 is titled "Graeco-Latin Alphabetical Glossary of Conjugated
Verbs" and its edition holds 576 Latin letters against 448 Greek ones
(ῥάπτω consuo, σπείρω spargo, κινέω moveo), yet <langUsage>
declares Greek only. Added <language ident="la">Latin</language>.

Noticed but deliberately left alone

So that these do not look overlooked: TM 171906 declares no en in
<langUsage> although its neighbours do; TM 62941 and 62944 carry
xml:lang="grc" on Ulpian's Latin name, which stays visible once the
@ref is gone. Both are outside what this batch claims to fix.

Validation

All 23 files validate against Validation/EpiDoc/8.23/tei-epidoc.rng with
Validation/jing.jar — the same schema and tool this repository's XML
Validation workflow uses for DCLP.

Every construction introduced is one the base already uses. Counted over
DCLP at 2249c22c92, before this batch: <author> without @ref — 9144
elements; <language ident="la"> — 1266; a negative notBefore — 1520.

One question

None of these commits touch <revisionDesc>. Almost every DCLP file
carries <change> entries with @who pointing at a papyri.info user, and
those are written by the Papyrological Editor's workflow, which a GitHub
pull request does not go through — #649 likewise merged without any. Should
I add a <change> per record, and if so under which identity, or is that
something finalization handles?

Happy to split this into smaller pull requests, or to withdraw any group
that reflects a convention I have misread.