Dependability means preserving useful work through change

Our standard for localization infrastructure includes the guidance, accepted translations, and operating knowledge a team will need after its next change.

A localization workflow becomes more valuable as a team uses it. Reviewers resolve ambiguities. Product experts refine terminology. Authors learn which instructions help translations preserve their meaning. That accumulated work deserves to remain useful when the source, the software, or the people involved change.

We treat continuity as part of the product relationship. A successful first request establishes a connection. Continued reliance requires understanding what has been accepted, what remains unfinished, and what the next change will affect.

Change the intended part of the work #

Changing a model, editing a shared glossary, and updating source content are different operations. A model change alters how later text is generated. A glossary change can affect every localization engine attached to it. A source change can make an earlier translation incomplete even if its wording was previously correct.

Useful infrastructure makes those consequences understandable. The team can then decide whether to preserve accepted output, regenerate a particular part, or compare the new configuration with earlier results. A saved configuration change should not be confused with a decision about every translation already in use.

Our responsibility includes the guidance held by the service. Terminology, voice, and instructions are part of what the customer depends on, alongside its source and accepted output. Explaining their scope and relationships matters when planning a change, not only when introducing the feature that stores them.

Recover the relationship between source and result #

After an interruption, finding a target file answers only part of the question. The team also needs to know which source it translates, whether review completed, and which work can safely resume. Repeating everything can spend more, overwrite useful corrections, or produce a new result without resolving what happened to the first one.

We judge recovery by the usable work it restores. Preserved bytes matter because they belong to a source revision, a language, and a decision about use. Resuming execution needs to respect those relationships. The interrupted-run guide explains how to investigate state before retrying.

Recovery coverage and commitments must be established in the relevant technical and contractual terms. Our standard does not make every resource recoverable by declaration. It requires a clear account of what the service preserves and what the customer must keep, so neither side discovers the missing responsibility during an interruption.

Let the next owner understand the workflow #

A colleague taking responsibility for localization needs more than working credentials. They need to know which system owns the source, which localization engine callers select, why the shared guidance exists, and where unfinished decisions appear.

The same applies to our part of the relationship. Product behavior and commitments should be understandable without depending on one person's memory of the original setup. Clear ownership and explainable configuration let the organization continue developing its language work as responsibilities change.

Make continued use a choice #

An organization may eventually change suppliers, content systems, or its own operating model. Translated files are one part of that transition. Terminology, instructions, locale mappings, review context, and delivery knowledge can also determine how much of the previous work remains useful.

We regard that transition as part of a dependable relationship. Obtaining material, understanding it in the next system, ending access, and settling retained records are separate responsibilities. Their scope should be clear enough for the customer to plan the change.

A provider earns continued use through the value of its contribution. Preserving the customer's ability to understand and reuse its work is consistent with that aim: each new translation can benefit from what the organization has already learned.