Format Autopsy
Run one file through a chain of format conversions, keep every stage, and argue at what point it stopped being the same file
Version 1.1 · Supersedes v1.0 · Status: CURRENT
- v1.0: Style pass only; no change to what is asked.
Due: 7 May 2027 · Weight: 25%
Set in week 10. Due 5pm Friday of week 11.
The brief#
Convert one file until it is no longer the same file. Say where that happened, and defend it.
Pick a source file with something to lose: comments, tracked changes, footnotes, a table, embedded fonts, non-Latin characters, transparency, layers, a colour profile. A plain paragraph of English text survives almost anything, which makes it a poor subject. Then run it through at least four conversions. The worked chain is DOCX to Google Docs to PDF to OCR to TXT to DOCX, but any chain that crosses a format boundary in both directions will do. Keep every intermediate artefact. They are part of the submission.
The week 10 lab did one step and itemised the losses. This does the chain, where the interesting behaviour is: losses that compound, a step that cannot tell an authored character from an artefact the previous step introduced, and a round trip that returns a file with the original extension, which opens in the original application, and fails every identity test from week 2 against the file you started with.
Record what changed at every step, and include changes that are not losses: a file that gets larger, a heading that becomes bold text of the same size, a straight quote that becomes a curly one. Then answer the question. At what point did it stop being the same file? Any step is defensible. The mark is on whether your verdict follows from the artefacts in front of you, and on whether you separate what the target format could not hold from what the converter chose not to write.
What you submit#
The artefact from every stage of the chain, named so the order is unambiguous, plus one document containing the per-step record and the verdict. State the exact tool and version used for each conversion. Two DOCX-to-PDF converters do not lose the same things, and the record is not reproducible without knowing which one ran.
The spec
The brief poses the problem and leaves room for a response. The spec is the fixed part: what a marker checks. Some lines can be checked mechanically; the rest call for judgement.
- a source file of your choosing, with a stated reason it is a useful subject
- a chain of at least four conversions, with the artefact from every stage submitted
- an itemised record of what changed at each step, including changes that are not losses
- a verdict naming the step at which it stopped being the same file, argued from the artefacts
How it is marked
There is no correct answer to where the file stopped being itself. The mark is for the argument, whether the verdict follows from the artefacts you produced, whether the losses you claim are ones you can point to, and whether you distinguish what the format could not carry from what the tool declined to carry. A defended early verdict and a defended late one can both score full marks. An undefended one cannot.