rte-06 reproduction
Public source retained with the related article. Generated results, dependency directories, runtime storage, secrets, and oversized binary artifacts are intentionally excluded.
README
# RTE-06: payload size is not the whole cache-hit cost
Run with Node.js:
```sh
node experiment.mjs
```
The model uses only standard-library assertions and gzip. It needs no external
service, application, dependency installation, credential, or file write.
For a fixed set of 800 synthetic rows, it prints expanded/tuple JSON byte
lengths before and after gzip level 1. It then compares eager expansion of all
800 rows with expansion of five selected rows. Both paths decode the complete
tuple payload first: selective expansion is not selective JSON parsing.
Assertions verify selected values including decimal strings and empty arrays,
field-descriptor reordering, and rejection of unknown formats, missing fields,
and wrong tuple lengths. The fixture is not a general schema validator or a
copy of the historical codecs. It omits the application's null/absence rules,
PHP serialization, Redis transport, connection waits, numeric computation,
and evaluator/materializer behavior.
Byte lengths may vary with compression implementation/version. No historical
payload size or compression ratio is asserted. This is not a latency, memory,
load, or cross-language benchmark. All data is synthetic public teaching
material, independent of unpublished article text and private research.