H.10 Platform

A large repository is served by SCM hosting, review, search, indexes, ownership, workspace, mirrors, snapshots, and other capabilities. Treating that portfolio as one invisible “repo” makes reliability, capacity, support, and cost impossible to reason about.

Define The Service Before Its Objective

H.10.1 Repository Services inventories clients, capabilities, authorities, data, dependencies, owners, and normal, degraded, emergency, maintenance, and retirement modes. H.10.2 Service Reliability can then set availability, integrity, durability, recovery, and data-loss objectives for named workflows. A restore exercise, not the existence of a backup, is the evidence that source metadata, review state, ownership, or indexes can recover.¹

H.10.3 Service Capacity segments cold and warm workflows, teams, languages, regions, change classes, partial workspaces, peak demand, backlog, growth, and failure state. Repository size alone cannot reveal whether transfer, indexing, query, review, concurrency, storage, or a downstream service is saturated.

Operate Demand And Economics

H.10.4 Platform Support routes incidents, product defects, documentation gaps, onboarding, feature intake, migration help, and domain questions differently. Repeated questions, workarounds, and exceptions feed product or policy repair; enabling work transfers capability instead of making a central queue permanent.

H.10.5 Economics compares build, buy, and hybrid services through engineering, infrastructure, storage, network, security, compliance, support, integration, incidents, migration, switching, and exit. The same option may move cost from developer wait to platform toil or from control to vendor dependency, so comparisons need workflow and cohort evidence.

H.10.6 Repository Metrics closes the loop with time to first change, feedback and review wait, recovery, repeated support, toil, and developer experience by cohort. These outcomes test whether service and policy changes improve repository work without collapsing people or workflows into one health score. Telemetry purpose, access, retention, aggregation, and privacy are part of that measurement contract.

Footnotes

1. The Site Reliability Workbook — Ch. 8: On-Call — explicit service responsibility, incident response, and recovery readiness ↩