This page answers a technical evaluator’s checklist. It lists the open standards and interfaces Backbone reads or emits, says which are shipped, which are in progress and which are scheduled, and gives a date for each. The same matrix is exported as a dated document for the trust center.
A checklist, answered one line at a time with a status and a date.
Evaluators and architecture review boards keep a checklist of the standards a data platform should support. This page answers that checklist line by line. Each line names the standard, says in one phrase what it is for, gives its status in one of four words, and gives a date.
The dates are of two kinds. A shipped line shows the version or the month it landed. Every other line shows a target quarter from the twelve-month plan that started in September 2026, so month 6 of the plan falls in Q1 2027 and month 12 in Q3 2027. The page is re-dated each quarter, and a target that moves is shown with its new quarter.
The same sixteen lines are exported as a dated document for the trust center, so a procurement or audit team can file the matrix as it is.
Available in the product today. The line shows the version or the month it landed.
9 linesEngineering work is under way. The line shows the target quarter.
3 lines · Q1 2027On the twelve-month plan. The line shows the target quarter.
3 lines · Q3 2027Being assessed by engineering. The line receives a date once a decision is made.
1 lineRead the table from top to bottom: the shipped lines first, then the lines in progress, then the scheduled lines and the one in evaluation.
| Standard | What it is | Status | Date |
|---|---|---|---|
| OpenAPI 3.1 | A machine-readable description of the platform’s REST API. It is generated from the running version, so the description matches the release you run. | Shipped | shipped |
| GraphQL | A query language for APIs. One request asks for exactly the fields it needs. Used beside REST for catalog, lineage and data-product queries. | Shipped | v2.15 · Feb 2026 |
| SAML 2.0 · SCIM | SAML gives single sign-on through Azure AD, Okta or LDAP. SCIM adds and removes users automatically from your identity system. | Shipped | shipped |
| Webhooks | Events pushed to your tools as they happen: quality findings, deadline risks, certification changes and publish events, to Slack, Teams, PagerDuty and ServiceNow. | Shipped | shipped |
| SDKsPython · JavaScript | Libraries for calling the platform from your own code. The connector SDK, for building your own connectors, is GA in Python, JavaScript and Java and in beta in Scala. | Shipped | shipped |
| Kubernetes (EKS) · GitOps promotion | The platform runs on Kubernetes. Configuration is promoted from development to staging to production with approval gates and a one-step rollback. | Shipped | shipped |
| Trino SQL dialect | The SQL dialect of the Trino query engine, used for questions across systems. The editor shows the plan and a dry-run cost before the query runs. | Shipped | shipped |
| Kerberos | The authentication protocol of Hadoop systems, for HDFS and other Kerberised sources. | Shipped | shipped |
| OpenLineage metadata sync | Lima’s datasets, jobs and lineage published to your catalog through Apache Atlas OMRS and OpenLineage. | Shipped | shipped |
| OpenLineage run events per task | Start, complete and hold events for every task, sent to your catalog or monitoring tool as a run proceeds. | In progress | Q1 2027 · month 6 |
| Apache Airflow 3 | A formal support statement for the Airflow 3 line, validated on Airflow 3.1. The draft is below. | In progress | Q1 2027 · month 6 |
| MCP serverread-only | An interface that lets AI assistants read catalog metadata. Row data stays inside the platform. | In progress | Q1 2027 · month 6 |
| Open Data Contract StandardODCS | Certified views and data products published as data contracts: the columns, their types, the freshness and the owner, in a form other tools read. | Scheduled | Q3 2027 · month 12 |
| Iceberg / Delta output | Lakehouse landings written as open table formats, so any query engine can read the same table. | Scheduled | Q3 2027 · month 12 |
| dbt interoperability | dbt models referenced in Lima lineage and scheduled inside the same run window as the pipeline that lands their input. | Scheduled | Q3 2027 · month 12 |
| Apache Arrow / ADBC | A columnar memory format (Arrow) and a database connection standard built on it (ADBC), for high-volume reads by analytical tools. | In evaluation | undated |
Drafted, published as a draft, and marked pending engineering sign-off.
Backbone’s orchestration runs on Apache Airflow, the open-source scheduler most data teams already use. Backbone runs on the Airflow 3 line today. The formal support statement is the document an evaluator asks for, and it has three parts: the Airflow line Backbone runs on, the version it was validated against, and the scope of the claim.
The statement is drafted. It says that Backbone runs on the Airflow 3 line and that support is validated on Airflow 3.1, including its human-in-the-loop operators and its deadline alerts. Engineering has the draft for sign-off, and the signed version is targeted for Q1 2027. Until then the draft is published as a draft, so an evaluator can read the exact wording that will be claimed.
The holds, the deadlines on the business calendar and the checkpoint resume described on the Orchestration page are Lima’s own layer above Airflow. They work on the Airflow line you run today.
Backbone’s orchestration runs on Apache Airflow. Lima pipelines appear as tasks inside Airflow DAGs. A DAG is a flow chart of tasks and the order they run in. One DAG can hold Lima tasks and your own tasks under one retry policy and one run dashboard.
The statement is validated against Airflow 3.1, including its human-in-the-loop operators and its deadline alerts. A human-in-the-loop operator lets a task wait for a person’s decision. A deadline alert raises a warning when a run is projected to miss its deadline. Both map onto Backbone’s holds, where a run waits for a person to fix the data, and onto its deadlines on the business calendar.
The statement covers the Airflow line Backbone ships with and the operators Lima provides. Customers who run their own Airflow keep their deployment and add Lima tasks to it. The statement is reviewed with every Backbone release and re-dated each quarter.
Signed: — · Dated: — · Review: quarterly · Target for the signed version: Q1 2027 (month 6 of the plan)
Lineage is a record of where each dataset came from and which steps produced it. A data contract is a written agreement on a dataset: its columns, their types, how fresh the data is, and who owns it. Both exist inside Lima today. The matrix is about publishing them in standard forms that other tools read.
OpenLineage is the standard for exchanging lineage between tools. Lima emits it today for metadata sync, through Apache Atlas OMRS: your catalog sees Lima’s datasets, jobs and lineage. By Q1 2027 Lima also emits a run event per task, with a start, a complete and a hold event, so your monitoring tool sees each run as it proceeds.
The Open Data Contract Standard (ODCS) is the standard form for a data contract. By Q3 2027 a certified view or a data product can be published as an ODCS document. A consumer can then check the contract with their own tools and hold the producer to it. Inside Lima the certified view and its tests do this today: a view that fails a test goes back to Draft, and every dashboard that reads it shows the status in place of a number. How tests guard a certified view →
Lima’s datasets, jobs and lineage reach your catalog through Apache Atlas OMRS and OpenLineage.
A start, a complete and a hold event for every task, so your monitoring tool sees each run as it proceeds.
A certified view or data product published as a contract: columns, types, freshness and owner, in a form a consumer’s tools read.
Four interfaces, three shipped and one in progress.
Other tools read Lima’s metadata through four interfaces. The REST API is described by an OpenAPI 3.1 specification that is generated from the running version, so the specification matches the release you run. GraphQL sits beside it for queries that want a chosen set of fields in one request. Webhooks push events to your tools as they happen. The MCP server, in progress for Q1 2027, is a read-only interface that lets AI assistants read catalog metadata; row data stays inside the platform. The specification and the webhook catalog ship with the standards matrix from the trust center.
| Interface | What it is | What it gives you | Status |
|---|---|---|---|
| REST APIOpenAPI 3.1 | The platform’s web API, described by a machine-readable specification generated from the running version. | Catalog, lineage, pipelines, runs and data products from your own code, with scoped API keys and OAuth. | Shipped |
| GraphQL | A query language for APIs where one request names the fields it wants. | Catalog, lineage and data-product queries in one round trip. | Shipped |
| Webhooks | Events pushed to a URL you register, as they happen. | Quality, deadline, certification and publish events in Slack, Teams, PagerDuty or ServiceNow. | Shipped |
| MCP serverread-only | An interface that lets AI assistants read catalog metadata. | Catalog entries, lineage and run state for an assistant. Row data stays inside the platform. | In progress |
The interfaces used on each ERP, side by side.
An ERP is the system that holds orders, stock, materials and finance, and each vendor sets rules for how other software may read and write it. The table lists the interfaces Backbone uses on SAP and on Oracle for reading tables, reading changed rows and writing results back, and the vendor rules it respects.
| SAP ECC and S/4HANA | Oracle E-Business Suite, Oracle ERP and Oracle Cloud | |
|---|---|---|
| Reading tables | Released function modules, BAPIs and CDS views. A BAPI is a function SAP publishes for reading or writing one business object. A CDS view is a database view SAP defines and releases. | JDBC database reads, the standard database connection, with a large table split into partitions so several threads read it together. A read-only database user is used. |
| Reading changed rows | Application-level change capture: change pointers, table-level change capture, delta-enabled CDS views and extractor delta queues. IDocs, SAP’s document format for exchanging business data, are read where a process already produces them. | Deltas by a last-update column on each table. LogMiner change capture, which reads Oracle’s own log of changes, where the database administrator allows it. |
| Rules respected | SAP Notes 3255746, 382318 and 2971304. The ODP-RFC interface, RFC_READ_TABLE in production and HANA database log capture are left out because those Notes restrict them. | Reads use a standard read-only database user with grants on the tables in scope. The database keeps its standard setup. Oracle Cloud is read through its released REST APIs and extracts. |
| Writing results back | BAPI and IDoc transactions, for example the released BAPI for a planned order. The ERP’s confirmation is captured for every write. | Open interface tables and public APIs for E-Business Suite and Oracle ERP; REST APIs for Oracle Cloud. The confirmation is captured for every write. |
ERP extraction, method by method → · Write-back and confirmation →
Parquet today; Iceberg and Delta tables, and dbt, by Q3 2027.
Backbone lands data where you want it kept, and the format it lands in decides which tools can read it afterwards. The table lists the file and table formats and the tools they work with, what is available today, and what is targeted. The ETL, ELT and mixed patterns are compared on the Pipelines page.
| Format or tool | What it is | Today | Target |
|---|---|---|---|
| Parquet | A compressed, column-based file format for analytical data. | Lakehouse landings are written as Parquet, CSV or JSON files to S3, Azure Data Lake or GCS, partitioned by date. | Shipped |
| Iceberg / Delta | Open table formats. A layer over Parquet files that adds a table with versions, so any query engine reads the same table. | Snowflake is loaded natively; Databricks Delta, Redshift, Synapse and BigQuery through their connectors. | Scheduled Q3 2027: landings written as Iceberg or Delta tables. |
| dbt | A tool modelling teams use to write transforms in SQL inside the warehouse. | The ELT pattern lands raw data and hands the transforms to the warehouse, where dbt runs them in the same schedule. | Scheduled Q3 2027: dbt models in Lima lineage and in the same run window. |
| Trino SQL dialect | The SQL dialect of the query engine used for questions across systems. | The SQL editor shows the plan and a dry-run cost before the query runs, and pushes filters down to each source. | Shipped |
| Arrow / ADBC | Arrow is a memory format for columnar data. ADBC is a database connection standard built on it, for analytical tools reading at volume. | Being assessed by engineering. | In evaluation |
Twelve capabilities and five use cases, one page each.
Analysts score data-integration tools on twelve critical capabilities and five use cases (Gartner’s framework). Each one below has a one-page answer with labelled evidence, available on request from the trust center. The lists give the plain meaning of each item.
Standards apply at every step: lineage as data is read and written, contracts on what is loaded, Airflow around the orchestration, open formats on the way out, and confirmed writes on the way back.
Yes, today. Metadata sync through Apache Atlas OMRS and OpenLineage carries Lima’s datasets, jobs and lineage to the catalog you already run. Run events per task, with a start, a complete and a hold event, follow in Q1 2027.
Backbone runs on the Apache Airflow 3 line, and Lima tasks run inside Airflow 3 DAGs. The formal support statement, validated on Airflow 3.1’s human-in-the-loop operators and deadline alerts, is drafted on this page and is pending engineering sign-off. The signed version is targeted for Q1 2027.
Inside Lima, yes, today: a certified view has tests, and a view that fails a test goes back to Draft before any dashboard shows a wrong number. From Q3 2027 the same terms can be published as an Open Data Contract Standard document, so a consumer can check the contract with their own tools.
Parquet, CSV and JSON files today, written to S3, Azure Data Lake or GCS and partitioned by date. Snowflake is loaded natively, and Databricks Delta, Redshift, Synapse and BigQuery through their connectors. Output as Iceberg or Delta tables is targeted for Q3 2027.
The platform API is described by an OpenAPI 3.1 specification generated from the running version, so the specification matches the release you run. GraphQL sits beside it, SDKs in Python and JavaScript wrap both, and webhooks push events out. The specification is part of the standards matrix export, so an evaluator can check the coverage of each screen against it.
The sixteen lines as a dated document, with the Airflow 3 draft statement, the OpenLineage mapping and the RFP answer library, from the trust center.
