Function parse_archive_listing 

pub fn parse_archive_listing(
    body: &str,
) -> Result<Vec<PublishedObject>, DataCatalogError>

Parse the object entries out of an archive listing body.

Dialect detection is closed: the body must classify as exactly one of the listing surfaces the catalog’s archives actually serve, each verified live on 2026-08-04 and recorded as a fixture, and a body that fits none of them is DataCatalogError::UnrecognizedArchiveListing - never a best-effort empty result. An error page, a login interstitial, or a format change at an archive must surface as “this is not a listing I understand”, because a silent empty parse is indistinguishable from “nothing published” and would convert an archive change into a false publication-gap report.

• Apache <pre> autoindex and its older table flavor (GFZ isdc-data.gfz.de, BKG igs.bkg.bund.de) and the ESA XHTML table autoindex (navigation-office.esa.int): recognized by the autoindex Index of marker; objects are relative anchors, with the row’s YYYY-MM-DD HH:MM text captured verbatim.
• AIUB whole-tree CSV (www.aiub.unibe.ch/download/full_listing.csv): path;bytes;ISO-8601;md5 rows; every non-empty row must fit that grammar.
• Anonymous-FTP LIST output (WHU igs.gnsswhu.cn): Unix ls -l rows (an optional leading total line allowed); every other non-empty row must fit that grammar.

Within a recognized dialect, rows that by the dialect’s own rules do not name an object (parent links, sort links, directories, symlinks) are skipped. The result preserves nothing but object paths and verbatim modification text; interpretation belongs to newest_published_product.