Impressum / Publisher information
PreviousNEO research archive and analysis website.
Publisher and contact
Jost JahnSmaeswai 4
25946 Nebel
Germany
PreviousNEO@jostjahn.de
Amrum Software is a project and collective name for the software offerings of Jost Jahn.
Publication workflow and measured runtimes (English)
Bounded serial publication
One owned Python worker executes MPC, data, enrichment, website, publication and HTTPS sequentially. A read-only Windows monitor checks its existing evidence every 15 minutes and starts no processing.
Normal ceiling: 45 minutes. At most 120 minutes only when the same five-second native sample proves CPU load of at least 85%, or at least 95% activity and 100 ms average read latency on a bound source disk. The initial decision never extends a saved running deadline. From the first admitted native process start. Restarting a process never resets a stage, freshness or total deadline. Exhausted attempts retain their evidence and do not automatically create another attempt.
After 45 minutes the controller requests a checked pause. An already running bounded unit may finish its small durable commit. The 120-minute outer task and the original stage and total deadlines still apply.
The complete regression suite is run for a program release. Publication may reuse only the activated result for the exact tested source and runtime bytes. Real input, transfer, TLS and complete-publication checks remain mandatory.
- MPC source snapshot
Fetch and verify the complete source pair, or reuse the exact retained snapshot.
Stage ceiling: 10 minutes, clipped to the original total deadline. - Scientific database and reports
Maintain the checked source index, reuse unchanged source-local observations, and update affected values and content-changed reports. Directory journal gaps require complete discovery; timestamps alone never approve reuse.
Stage ceiling: 120 minutes, clipped to the original total deadline. - External identification data
Preserve source provenance and update the required identification data.
Stage ceiling: 30 minutes, clipped to the original total deadline. - Website and original exports
Reuse verified original exports and render content changes; retain unchanged data and files.
Stage ceiling: 60 minutes, clipped to the original total deadline. - Changed-content publication
Check changed local contents and retain exact successful target hashes. The central uploader verifies new transfers before accepting them.
Stage ceiling: 120 minutes, clipped to the original total deadline. - Changed and critical HTTPS checks
Verify changed files and critical entry pages with full HTTPS bytes and SHA-256. Preserve successful unchanged receipts with their original dates. Resume only exact atomic partial receipts within the original fixed window.
Stage ceiling: 120 minutes, clipped to the original total deadline.
Only changed local contents need another preparatory full-byte read. Unchanged targets retain their exact destination/content proof. A missing or inconsistent proof requires validation; modification time and size alone never prove unchanged content. The central upload gates remain in force.
Public HTTPS scopes: object files at jostjahn.de/neocp-obs80/, website files at jostjahn.de/previous/, and the website alias at artsat.jostjahn.de. New uploads receive complete checks before acceptance. Unchanged files retain their exact historical destination/content receipts; they are not reported as fresh downloads.
Local hash reuse requires a protected full read followed by the same native file and volume identity and continuous NTFS journal proof. A zero file USN additionally requires complete journal replay. Missing access, unsupported records, a changed file or a journal gap causes a full read. Modification time and size alone never authorize reuse.
Latest retained measurements
These are dated observations of the retained attempt, not promises or a live monitor. Previously completed calculations are reused and are not charged as newly executed calculations.
- MPC source snapshot
- 6.953 seconds
Measured worker time across completed attempts. Wall-clock span through the last saved attempt 7.0 s; 0.0 s outside measured worker intervals. - Scientific database and reports
- 129.266 seconds
Measured worker time across completed attempts. - External identification data
- 31.969 seconds
Measured worker time across completed attempts. - Website and original exports
- 75.219 seconds
Measured worker time across completed attempts.
Retained attempt status: running. Source dataset: 2026-11-17.53.
No complete recent website check is recorded in this disclosure. This check does not turn a failed production attempt into a successful one or make retained source data newer.
The diagram is regenerated from the controller's actual stage configuration whenever the website is built. Unavailable measurements remain explicitly unavailable.