Why everything here is cited

Every saying on this site names the source its wording came from, a dictionary or archive where scholarship allows, a stated published use where it does not yet, with the page saying plainly which kind you are looking at. That rule sounds like housekeeping. In practice it is the entire product. And the interesting part is not the rule but the traps we hit while following it. Copying Irish carefully out of real sources turns out to be genuinely difficult, in ways that explain how so much corrupted Irish got loose in the first place.

Trap one: the dictionary abbreviates itself

Open Ó Dónaill’s Foclóir Gaeilge–Béarla (the standard modern dictionary) and the example phrases under a headword do not repeat the headword. They abbreviate it to a tilde. The entry that attests our record Go dtí tú slán literally prints Go ~ tú slán: to know what the blessing says, you must expand the tilde with the exact headword form the entry sits under. And headwords have grammatical forms, so expanding carelessly can produce a “dictionary-attested” phrase the dictionary never printed.

This one bit us productively. The phrase sits under a literary verb form meaning come (dtí) one letter away from the subjunctive of the verb go, and the dictionary’s own gloss, “may you come back safe”, is what settles which verb the blessing uses. One letter, one tilde, and the difference between a blessing for the return and a blessing for the departure. Expanding an abbreviation is an editorial act, so our citation records the expansion: the locator on that record quotes the entry tilde and all, then states what we did.

Trap two: the official source with a typo

The Government of Ireland publishes a list of eleven seanfhocail. As close to an official imprimatur as an Irish proverb gets, and a source we happily cite. Mining it, we found one of the eleven misspelt: a proverb about high and low station rendered with a non-word where uasal, noble, should be. uasak, k and l being keyboard neighbours. Ten entries were perfect. The eleventh would have shipped a misspelt Irish word under a government citation to anyone copying on autopilot.

The lesson is precise: “as printed” cannot mean “copied blindly”. A source attests a saying; it does not sanctify its own typos. When a source’s wording looks wrong we treat that as a finding to record, not a licence to silently correct. That proverb has no record here yet, and the typo is documented in our files so the next person mining that page knows.

Trap three: the source is more faithful than your character set

Our validation rejects any character outside the Irish alphabet and its ordinary punctuation: that is how we catch the counterfeit accents that render like a fada and are not one. Then Ó Dónaill attested a proverb with a parenthesis in it: Tá Dia láidir (agus tá máthair mhaith aige). God is strong (and he has a good mother). The parentheses are the dictionary’s own; the aside is the proverb’s whole charm. Our rules banned them, which left two options: truncate sourced text, or refuse it. Truncating sourced text is the exact crime the rules exist to prevent. So the character set gave way, in a reviewed change with a test pinning the decision. Rules serve the sources, not the other way round.

Trap four: the echo chamber

The reason quote sites disagree with the dictionaries is that most of them cite each other. A wording is invented or mangled once, copied a hundred times, and the copies then look like corroboration. That is why our method refuses agreement between quote aggregators as evidence of anything, and refuses agreement with our own other pages hardest of all. A second copy of our data is a copy, not a confirmation. When two of our genuine sources disagree about a wording, both wordings go into the record as attested variants, each with its citation; choosing silently between sources is just inventing with extra steps.

What the machines check, and what they cannot

Every page of this site is built from record files, and every build runs the record store and the rendered pages through validation: schema, cross-references, citation presence, and fada-lint on both the stored and the rendered Irish: down to parsing our own font files to prove the accented letters are still in them. A saying without a citation cannot ship; the build fails instead. Even the essays you are reading are policed: a machine check refuses any Irish phrase in an article that is not one of the published, cited records, which is why every Irish word on this page is a link.

What no machine can check is the step where a human reads Go ~ tú slán in a dictionary and types the expansion. So that step is governed instead: changes to a record’s Irish or its citations get human review, and a wording in wide genuine use whose earliest attestation we have not pinned down is published with that fact printed on its page. Cited to a named published use, never to nothing. The remaining queue of open questions is public in our issue tracker rather than quietly shipped. Even Sláinte!: one word, known to everyone on earth: carries its dictionary citation here, because a promise with one exception is a different and weaker promise.

The result is a smaller site than the aggregators. It is meant to be. Someone about to carve five Irish words into stone does not need ten thousand sayings; they need five words that are right, and a way to see for themselves where each came from. That is the whole method: show the shelf, name the page, and let the reader check us.