What is missing
On the conformance map every section is marked “partial”. Here is what stands behind that word: 16 gaps across 9 sections.
The list is written without regard for how it looks. A specialist who opens the system will find what was passed over here anyway — and after that will believe nothing else either. Knowing the boundaries in ten minutes is a better deal for both sides than learning them in the third month of a rollout.
General Rules
Section 01 ↗General rules of recording: who keeps the records, how their trustworthiness is confirmed, what a recording organisation does and what it answers for before the breeder.
How it stands now. Roles, access rights and a change log on every record. Verification of records by the Association runs as a separate stream of applications.
There is no external audit of the procedures.
What it costs. A rule checked only by whoever wrote it is a promise, not a rule. The section is built around trustworthiness being confirmed by a third party, and a run of our own cannot take its place.
What it would take. The ICAR Certificate of Quality for Herd-book recording and Data processing — the only industry certification of herd book keeping in the world. It is open to ICAR members only, and that runs into matters outside the code (see docs/icar.md).
The recording procedures are not written down as a document of their own.
What it costs. The order of work lives in the code and in the decisions of the developer. While one person runs the system that works; on handover to the Association, or under an external audit, it does not.
What it would take. Written rules for keeping the book: who enters, who checks, within what deadlines, what counts as an error and how it is corrected. Work for the Association, not for the system.
Cattle Milk Recording
Section 02 ↗Recording of milk yield: test-day schemes, the intervals between them, computing the lactation, the designation of the performance recording method (A4, B4 and the rest) and the handling of missing tests.
How it stands now. Test days with date, yield, fat, protein and somatic cells; the lactation is computed from them rather than entered as a number. The performance recording method is written down through the enumerations of the standard: who took the readings, which milkings were included, how and when the sample was taken — and the familiar A4 or B4 designation is assembled from those rather than stored as a string. Checks catch breaks in the run of test days and lactations without the mandatory figures.
The interval between test days is not checked.
What it costs. The section sets the permissible intervals, and going beyond them means the lactation has been computed from points too far apart. Our check catches only a complete break in the run, not a stretched interval.
What it would take. A rule in the check registry: the interval between neighbouring test days, in days.
DNA Technology
Section 04 ↗Work with genotypes: parentage verification and discovery, the exchange of genotypes between organisations, and the requirements on laboratories.
How it stands now. A DNA test with the laboratory, date, method and the result of parentage verification; a disagreement with the recorded parents goes into the checks.
The ISAG STR panel is stored, but not chip genotypes.
What it costs. Twelve loci are enough to confirm parentage and not enough for either genomic evaluation or the exchange of genotypes: there the count runs to tens of thousands of SNP. Chip files the book does not accept at all, and on a change of laboratory what remains of the earlier genotyping is a report, not the data.
What it would take. The twelve mandatory STR are already stored; chip genotypes take six steps, and not one of them is skipped. First — intake of the three formats they are handed over in: an Illumina chip report together with the panel manifest, a pair of PLINK files, and VCF with a declared genome assembly. Second — reduction to a single allele coding and a single assembly: one and the same genotype looks different in TOP, FORWARD and AB, and marker positions in ARS-UCD 1.2 and 2.0 do not agree. Third — a common reference panel and imputation up to it, otherwise animals from different chips are not comparable with one another. Fourth — the link between sample and animal: the sample identifier at the laboratory is not the animal number, and without a genotyping request the result hangs unattached. Fifth — quality control: the call rate for the sample and for the marker, the minor allele frequency, a sex check and a check against the recorded pedigree. Sixth — storage: fifty thousand markers per animal do not live as rows in a database; what is needed is a file in a binary format with metadata in the book — animal, laboratory, panel, date, number of markers, call rate and the file checksum. Only after that are runs of homozygosity, genomic relationship and exclusion probabilities computed. The ADE standard is no help here — it has no genotypes at all.
There is no exchange of genotypes with other organisations.
What it costs. International exchange goes through GenoEx-PSE, and it is closed: ICAR membership plus certification as a DNA Data Interpretation Centre are required. No Russian organisation is among the twenty-seven centres.
What it would take. Genotype storage of our own first; exchange becomes possible only once the external situation changes.
Conformation Recording
Section 05 ↗Conformation recording: linear traits on a nine-point scale, composite scores on a hundred-point scale, and the requirements on the classifier and on the repeatability of scores.
How it stands now. Linear scoring on traits 1–9, composite scores 50–100 and the scoring of young stock. Every score carries the classifier and the date. Sixteen linear traits and four composites are mapped to the ICAR nomenclature when exchanging over ADE.
Two composite traits do not fit the ICAR nomenclature.
What it costs. “Body volume” and “rear third of the body” are composites of our own, and the icarConformationTraitType enumeration has no counterpart for them. In an exchange they simply do not travel. Passing them off as the nearest traits by meaning would be sending another system a figure under someone else’s name.
What it would take. Either a decision by the Association to move to the ICAR set of traits, or the recognition that these two remain internal. The decision is a breeding one, not a technical one.
The repeatability of scores between classifiers is not measured.
What it costs. The section requires that different classifiers give one animal close scores and that this is measured. Without such a check the conformation rating reflects the habits of a particular classifier as much as it reflects the cow.
What it would take. Repeat scoring of the same animal by different classifiers and computation of the divergence. That data has to be collected on purpose — ordinary work does not produce it.
AI and ET Data and Fertility Analysis
Section 06 ↗Recording of inseminations and embryo transfers, and the computation of fertility figures: days open, the calving interval, services per conception.
How it stands now. Inseminations with the sire, dose, technician and result; days open and the calving interval are computed from the events. Checks catch a calving interval that is too short and an insemination dated before the calving.
Embryo transfer is not recorded.
What it costs. The section covers ET as well, and we have neither a flushing, nor a donor and a recipient, nor documentation for an embryo. A calf from a transfer is recorded as an ordinary one, and its pedigree comes out wrong: the recipient dam enters it instead of the genetic dam.
What it would take. A flushing event, an embryo record and a “recipient” relation type in the pedigree. ADE has everything for this: icarReproEmbryoFlushingEventResource, icarReproEmbryoResource and the value Recipient in icarAnimalRelationType. Required by the WHFF rules on registration.
Bovine Functional Traits
Section 07 ↗Functional traits: udder health, calving ease, calf survival, longevity — what exactly to record and how to code it.
How it stands now. Somatic cells from the test days, calving ease, live and dead calves, and disposal with the date and the reason.
The ICAR Central Health Key is not used.
What it costs. Disease records are kept more freely than the reference list requires, and the frequency of mastitis here cannot be compared with another country: there it is a code, here it is text. The same closes the road to international evaluation of health traits.
What it would take. The Central Health Key reference list, a code field on the treatment and the diagnosis, and conversion of the existing records. ADE uses the same key in icarDiagnosisEventResource.
Calving ease in three degrees instead of five.
What it costs. We have “easy, with assistance, difficult”, while the international scale has five degrees with a caesarean of its own. In an exchange our three map onto the other five unambiguously, but not back: two different foreign values arrive as one of ours, and the difference is lost silently.
What it would take. Extend the reference list to five degrees per icarReproCalvingEaseType and convert the existing records.
Dairy Cattle Genetic Evaluation
Section 09 ↗Genetic evaluation of dairy cattle: what breeding value is made of, how reliability is computed, how results are published, and the comparison base.
How it stands now. A breeding value index with weight profiles, reliability for each trait and for the index as a whole, and the version of the comparison base beside the value. The ranking within the book is published by name.
The comparison base is borrowed rather than our own.
What it costs. Standard deviations and heritabilities are taken from the American CDCB-2025 base and converted into metric units. For the Russian population they are approximate: the index is internally consistent, but its absolute value does not mean what it means in a country with a base of its own.
What it would take. Computation of genetic parameters on the Russian population — work for a research institute, not for the system. Our part is the base version beside every value, so that a change stays traceable; that is already done.
There is no validation of the genetic trend.
What it costs. The section requires checking that the evaluation does not drift from year to year: Interbull methods I–III are about exactly that. Without such a check a slow shift of the base looks like genetic progress, and the substitution cannot be noticed.
What it would take. Implementation of the trend validation tests, methods I–III. It is done without any membership and remains the right architecture even if participation in Interbull never happens.
The evaluation takes no part in international comparison.
What it costs. Interbull MACE brings the evaluations of different countries onto a common scale. Without it our index is comparable with no one’s in either direction — neither to compare a sire nor to judge an imported cow.
What it would take. Requires ICAR membership and a validated national evaluation in the common base. Closed today.
Data Exchange
Section 15 ↗Data exchange between systems: dictionaries, formats, schemas. The current line is the open ADE standard on JSON and REST, its specification held on GitHub under Apache 2.0.
How it stands now. Exchange with FGIAS PR over the twenty templates of the registry — export and reverse import. A REST interface of the book’s own. 14 ADE collections served at the paths of the specification (animals, test-day-results, parturitions, inseminations, type-classifications, weights, breeding-values, arrivals, departures, deaths, pregnancy-checks, milking-visits, daily-milking-averages, diagnoses). POST intake on 7 of them (test-day-results, parturitions, inseminations, weights, milking-visits, diagnoses, breeding-values) with repeat submissions recognised by the pair “source plus its record number”, a line-by-line icarBatchResult response and refusals as icarErrorResource. What is served is validated against the real schemas from the adewg/ICAR repository rather than against our own copy of the enumerations. Selection uses the names of the standard itself: meta-modified-from and meta-modified-to by record date, date-from and date-to by event date, the animal-id and animal-scheme pair by animal; parameters that were not understood are named in the response header rather than dropped silently. Animals, conformation scores, breeding values and movements are deliberately closed to writing: registering an animal and the transfer of rights are statements the Association answers for, and they go through an application with verification rather than as a line in an exchange stream; each is answered with 405 and an explanation.
The exchange carries 14 resources of the standard, and there are some fifty.
What it costs. Both ways of exchanging work in full, but they carry what the book keeps. Treatment, feeding, group events, slaughter, device readings are resources of the standard that we do not have in the book itself, and the exchange has nothing to do with it: there is nothing to carry. Treatment has a reason of its own, and the client is told it in the refusal: there is nowhere to keep the withdrawal period, and taking the record while losing it would be worse than not taking it.
What it would take. Keeping the corresponding sections in the book. The exchange will pick them up without a change to the protocol: the collection path and the data set are added in one line — which is exactly why the standard was made independent of the resource type.
Breed Associations
Section 18 ↗The work of breed associations and the keeping of herd books: the rules for entering an animal in the book, the sections of the book, and the requirements on descent and on membership.
How it stands now. Keeping the book, membership of holdings, issuing breeding certificates, handling applications, verification of records.
The herd book has no sections.
What it costs. In international practice the book is divided by degree of purity of breeding (a main part and appendices), and what an animal can be granted and the price it sells at depend on the section. Our book is single and flat.
What it would take. A decision by the Association on the structure of the sections, then a section field on the animal and the rules for entering it. The first matters more than the second and is not our work.
The pedigree does not guarantee five generations.
What it costs. The WHFF Registration Guidelines require five generations with performance data and a flag for a non-Holstein ancestor among them. We collect as much as the data holds and do not mark where the pedigree breaks off or where another breed enters it.
What it would take. A pedigree completeness attribute on the animal, a flag for a foreign breed within five generations, and codes for the Mulefoot, BLAD, CVM and DUMPS recessives.
What here does not depend on code
Some gaps are closed by work and some by a decision, and the two are worth keeping apart. The sections of the herd book, the set of conformation traits and the structure of performance recording methods are decisions for the Association: the system can do it either way, and the choice belongs to whoever answers for the breed. Computing genetic parameters on the Russian population is work for a research institute. International comparison of evaluations and certification of the quality of book keeping run into ICAR membership, and that is a separate conversation, in which little depends on us.