13 · Disclosure — what we have not done

A text that counts must also count what it does not know and has not done. The list below is complete, it is uncomfortable, and it is the reason the rest can be believed. It stands here rather than in a footnote, because a text that stays silent about its own starting position is worth less, and because each of the preceding twelve counts is worth only as much as our willingness to set our own beside them.

The sentence that carries all of it, unabridged:

We have never imported a file into a cantonal portal — neither an eCH-0276 file nor an XBRL instance document, in any canton. And we have never made a call to the swissdec Distributor, not even Ping.

Nothing in this text rests on a test; everything rests on documents. Where a statement about how a portal or an interface behaves would have been possible, the wording of the source stands instead. That is not a formula of modesty but the limit of the method: we read, we hashed, and we recounted; we pressed nothing.

What we built for — and which path is open today

The most uncomfortable line in this text is one about us. For Switzerland we built a generator for eCH-0276 V1.0.0 — that is, for the artefact that, in the communication matrix of the transmitter specification, goes out of a Deklarationslösung, not for the one that comes into it from an accounting system. The research for this text showed two things about that decision.

First: the path a private Deklarationslösung would have to take via the Distributor is the path of which the body that carries the network writes that «zuerst die vertraglichen Grundlagen geschaffen werden» (the contractual foundations would first have to be created). We built, in other words, for the lane that is contractually suspended. That is our risk, and it stands here because it speaks against us.

Second, and this is the correction to our own earlier version: the open path is not only XBRL out of the accounting system. The rulebook of the SSK, part of the same standard, names two recognized electronic submission channels for the dialogues Aktiven, Passiven, and Erfolgsrechnung, three times and in identical wording: «Wird die E-Bilanz elektronisch übermittelt (Upload oder via Swissdec), können die Positionen nicht überschrieben werden.» (if the E-Bilanz is transmitted electronically — upload or via Swissdec — the positions cannot be overwritten). Upload is a first-class channel within a component of the standard, and it requires no contract with swissdec. We wrote that too narrowly in an earlier version; the consequence for manufacturers stands in part 14.

What follows from both is not a product announcement but a placement: in Switzerland there is nothing to buy from us, and this text announces nothing. It is the evaluation of research we ran for our own decision.

The inventory list

No Zürich access credentials. The Registernummer and the Zugangscode go out by post, under § 4 Abs. 1 LS 631.121, to a Zürich company entered in the Steuerregister. We have none. The precondition for step 4 of our own test sequence is therefore absent, and our spec has said so, unchanged, for months: «I researched the channel; I have not tested a live Zürich connection — do not claim a working CH transmission until a real import test passes.» (CH-TRANSMISSION.md, test plan).

ebilanzfabrik.ch is not live. The A record still points to the registrar rather than to the intended server; what answers there is a parking page without a matching certificate. The CH deploy script has existed since 04.08.2026 and has never been run — no ssh, no DNS change, no request to the domain.

And here we correct ourselves in the act of writing. Our own register (CH-LANDING-OPEN-ITEMS.md, item 10) records that all eight Swiss pages carry <meta name="robots" content="noindex, follow">. Re-measured on 04.08.2026: six of eight carry it — the six legal pages. The two landing pages carry index, follow. In practice that changes nothing, because none of the eight pages is reachable as long as the A record stands. Methodologically it changes everything, and that is why it stands here: we did not carry the statement over from our own register; we opened the files.

No Swiss input path. No CH funnel, no upload UI, no template. The existing input path is tailored to § 266 HGB; the Swiss side is tailored to the OR. No CHF payment path: grep -rin '\bchf\b' across the web app returns zero; the paywall is pinned to an amount in EUR cents.

No Swiss legal presence of any kind. No Swiss company, no Swiss address, no UID, no MWST registration, no Steuervertreter under Art. 67 MWSTG, no representative under Art. 14 DSG. No proof of a working hello@ebilanzfabrik.ch mailbox — the only address with a documented receiving path ends in .de.

The provider is a German sole proprietorship; the full provider details are set out in the Impressum. No Swiss cross, no flag, no shield, no claim of Swiss origin. In Germany we are in production; in Switzerland nothing is for sale, and this text announces nothing.

The limits of the tooling

Neither Arelle nor lxml is installed on the system — both checked on 04.08.2026 by attempted import. The XBRL statement in this text is therefore a mapping, not a validated instance. On the eCH side, xsd_validate sits permanently in graceful-skip, because the official XSD imports eCH-0007, eCH-0010, and eCH-0129 without schemaLocation, and an offline build cannot resolve them. The gate that runs is the structural check of the standard library, not the schema.

What we asked, and what we have no answer to

The query to info@ech.ch — which artefacts a portal actually accepts, and what advice is given to a provider that has already implemented against V1.0.0 — sits in our review inbox and has not been sent. The enquiry to info@xbrl-ch.ch about the sample data and the entity-identifier scheme was not confirmed as sent as of 04.08.2026. Addendum of 05.08.2026: it was sent that day — with three questions, on the version and source of the taxonomy, on the relationship between instance document and transmission format, and on a means of testing before activation — and it was answered the same day. What the answer says stands here only once we may quote it; we have asked. Until then neither open question carries a statement in this text.

The extent of our own reach, which we control and therefore report where we would not report the state of others: _PILOT_CANTONS = {"ZH"} and _CH_LEGAL_FORMS = {"GmbH","AG","Verein","Stiftung"}.

Two retractions

The first, shown rather than described. Our note of 04.08.2026 opens with the sentence that the headline of its own predecessor — «kein kantonales Dokument nennt eCH-0276» (no cantonal document names eCH-0276) — is «nun durch zwei Primärquellen falsifiziert» (now falsified by two primary sources). Both notes carry the same date; the file timestamps are thirty-eight minutes apart. A headline published and retracted the same afternoon, against the same corpus, by the same author.

The second concerns this text. The statement in our manifest that the two standards do not reference one another at all is wrong — the reference linkbase of the CH-Taxonomie has carried eCH-0276 position codes under the label «Tax code» since 23.06.2024 (part 11). The count in the manifest was right; the conclusion drawn from it was not.

The distinction that decides the matter, verbatim from our own note:

«Unser Standardsatz deckt das Ausführungsrisiko ab (‹wir haben den Import nicht getestet›). Dies hier ist ein Existenzrisiko — ob der Weg für unser Artefakt überhaupt existiert. Das ist nicht dasselbe, und das zweite entscheidet über das Vorhaben.» (our standard sentence covers the execution risk — ‹we have not tested the import›. This one is an existence risk — whether the path for our artefact exists at all. That is not the same thing, and the second decides the undertaking.)

Errors in our own records

Recalculated before publication, all on 04.08.2026:

• The manifest dates the Zürich media release to 21.07.2025 and even carries that date in its file name. The document itself carries 18.07.2025 in line 55.
• The manifest says «10 SOAP samples» in two places. ls samples/*.xml | wc -l returns 11.
• For Schnittstelle* in ZStB 109c.1 the manifest names 19 in one place and 13 in another. Re-measuring with word boundaries returns 20; the total falls from 32 to 28 once the two renditions of the same Q&A are no longer counted twice.
• The manifest dates eCH-0276 V1.0.0 to 05.12.2024; the main document says «Beschluss am 2024-12-02» (decision on 2024-12-02) in line 19. 05.12. is the publication date.
• And the figure «64 Artefakte / 338'469 Wörter» (64 artefacts / 338'469 words) from our own briefing note does not reproduce: the tree holds 255 files, 91 MB, of which 57 are text extracts with 331'014 words.

The rule that follows: recalculate every number ourselves before publication; never carry one over from our own manifest. Five deviations in a bundle we wrote ourselves are not an argument against counting. They are the argument for counting twice.

The dating limit we do not cross

The widely circulated figure «produktiver Betrieb ab 2027» (productive operation from 2027) comes from swissdec.ch/ebilanz-info — a page that is neither crawled nor hashed. grep -rn "2027" across the corpus returns hits in exactly three files, and in none of them is 2027 attached to a swissdec production date. A text whose method is «gezählte Belege mit Prüfsumme» (counted evidence with a checksum) cannot carry its only forward-looking date on an unhashed page. The documented date therefore stands here: DVS-Vorhaben INM2.084, Leistungsverantwortliche Organisation SSK, Endtermin 31.12.2026, with two measures not marked as completed — M3 «Umsetzung in Pilotkantonen» (implementation in pilot cantons) and M5 «Umsetzung Pilot mit MWST-Verwaltung und Swissdec» (pilot implementation with the MWST administration and Swissdec) — and a budget line that shows zero for 2027. Anyone using the 2027 figure marks it as «swissdec.ch/ebilanz-info, abgerufen [Datum], ohne Offline-Beleg» (swissdec.ch/ebilanz-info, retrieved [date], without offline evidence).

What is missing from the corpus and therefore carries no statement

Nachtrag 111 of LS 631.121 — that is, the wording before 13.05.2025; the artefact fetched for it, zh/lexfind-631.121.txt, is not text but a corrupted PDF byte stream, and carries nothing. Further: the full texts of LS 631.122 and LS 631.43 · ABl 2025-05-23 / OS 80, 149 · §§ 109c, 109d, 133 StG in full text · the swissdec certification test cases [TFBASIS] · [SECPDF]/[SECRXPDF], both behind infopoint.swissdec.ch · the Institutionsliste retrievable on swissdec.ch under § 8.16 · Formular 500 exists only as a PDF and is not extracted; Formular 510 is missing entirely.

All of this stands here without the old rider that it is the only opinion in the text. This text is full of opinion about the situation — about what the documents mean, what happens next, and what has to change. That is the point. Opinion about the situation costs nothing but evidence. Opinion about one's own starting position costs something, and that is why it stands here unabridged.

---