Back to matrix

G05-TS: Pending

Rationale

G05 admits a connected multi-tailnet composition; one global tailnet is not required. Documented sharing gives access to selected resources through explicit bilateral policy or individual invitations. Connectivity of the sharing graph alone does not establish transitive endpoint reachability. Current declarative-sharing alpha limits (three external tailnets, 500 nodes each) constrain that mechanism, but neither define a permanent architectural ceiling nor exclude link-based sharing. The service-wide million-node observation lacks a per-tailnet hardware/resource curve. Conditional sizing separates daily accesses from newly created relationships; without locality, retention and update-fanout evidence, neither all-accesses-are-control-requests nor an assumed cache-hit rate is justified. A contemporary global placement remains unestablished, so G05 stays Pending.

Assessment method

Official sharing semantics and reproducible conditional arithmetic; verdict unchanged

Source context

Tailscale sharing documentation and service architecture; explicit sensitivity model

Outstanding qualification

Model one connected network using either one tailnet or interconnected tailnets, retaining managed coordination, Tailnet Lock and the common regional physical infrastructure. Cover state placement, direct cross-tailnet reachability, popular destinations and update fanout. Do not infer transitive routing from sharing, count tailnets as physical servers, require all possible pairs simultaneously, or assume every daily access creates a new relationship. The sizing artifact is conditional arithmetic, not a runtime result or capacity verdict. IPv6-only operation remains admissible; no denial based on the IPv4 /10 or centralization alone.

Public sources

Editorial evidence references

The following identifiers refer to research records not included in this public package; they are not public links.