Compressed vs Expanded Size Review: Record Resource Limits Before Reading an Archive

The source dates shown in this guide identify documentation checks on 22 October 2026, not original publication dates. Follow each citation for the container behavior being discussed. Start with a controlled artifact reference and record units explicitly. At minimum, distinguish bytes received, bytes stored in the wrapper, bytes declared for a member, bytes emitted after decoding, bytes actually inspected, and bytes retained in a derivative.

The Python Software Foundation, 22 Oct 2026 (opens in a new tab) documentation describes ZipInfo.file_size and compress_size as separate member facts. Python Software Foundation, 22 Oct 2026 (opens in a new tab) exposes tar member metadata, but a tar stream can contain padding, multiple members, and special types. A gzip wrapper has a different contract: RFC Editor, 22 Oct 2026 (opens in a new tab) defines ISIZE as the original input size modulo 2^32. Do not present one field as a universal expanded-size truth.

Attach a unit, source, and confidence to each value. “Unavailable” is different from zero, and a stopped counter is different from a complete count. Keep compressed and expanded artifact identifiers separate so a report never implies that a derivative is the original source.

Compare declared and observed values only after confirming that they describe the same member and representation. A mismatch can indicate truncation, a stale header, a concatenated stream, a reader policy, or an incorrect assumption about which layer was measured. It does not by itself identify cause or prove malicious behavior.

Synthetic size ledger

artifact_ref       layer                 declared_bytes  observed_bytes  state
synthetic-archive   wrapper               18432            18432           complete
synthetic-member   compressed member      5120             5120           complete
synthetic-member   expanded declaration  12000            12000           matched
synthetic-member   inspected window      unavailable      4096            bounded
synthetic-gzip     ISIZE trailer          modulo field     4294967295      not a full-size proof

The values are invented for this worksheet. The final row illustrates why a gzip ISIZE value must be labeled with its modulo semantics, not copied into a “total expanded bytes” column. The observation state explains whether a reader reached the end, stopped at a configured limit, or could not determine completion.

A useful budget is a policy record, not just a ratio. Define maximum container bytes, maximum member count, maximum expanded bytes, maximum nesting or wrapper depth, maximum retained bytes, and an external CPU or wall-clock budget. The application or execution environment must enforce those limits; a worksheet cannot make a decoder bounded after the fact.

1. 01Identify the archive and member before allocating a destination or output buffer.
2. 02Check available declarations and reject impossible totals before reading expanded data.
3. 03Count output incrementally and stop at the configured expanded-byte limit.
4. 04Drain or close the bounded stream according to the receiving runtime so the next member is not misread.
5. 05Record the stop class, counters, and cleanup result without storing raw expanded content.
6. 06Use a separate policy for nested archives and multiple members; a per-member limit is not a whole-job limit.

Start with text fixtures that exercise matched sizes, missing declarations, a member that reaches the expansion limit, a truncated stream, a concatenated wrapper, and a nested archive marker. Use synthetic references and labels. Do not copy a private archive name, path, object tag, or content sample into a public report.

When a limit is reached, keep the partial artifact out of any workflow that assumes completeness. If a bounded sample is useful, name it as a derivative, record its limit, and retain the source reference separately. A successful sample review is not evidence that the full expansion was observed.

Size and limit evidence can answer whether a selected run completed under a named policy. It cannot establish archive ownership, permission to retain content, privacy clearance, semantic accuracy, or commercial demand. Put those decisions in their own provenance, privacy, rights, and owner-approved records.

Pair this guide with the data provenance log for artifact identity and export guidance for producer behavior. After the owner has a supported non-health scope, authority to describe the collection, and a controlled size ledger or synthetic fixture, the qualified appraisal request path may be a next preparation step. It does not expand archives, certify resource limits, create a quote, match a buyer, issue a legal opinion, or grant a license.

Takeaways

• Record container, compressed, declared expanded, observed expanded, inspected, and retained sizes separately.
• Label units, source layers, completion state, and unavailable values explicitly.
• Treat gzip ISIZE as a modulo field rather than a universal expanded-size proof.
• Set per-member, whole-job, nesting, retention, and external runtime limits before expansion.
• Keep a limit stop or partial sample out of workflows that require complete evidence.