I'm reading "Requesting metadata with electronic records" from Christine Beckett, writing for the Reporters Committee for the Freedom of the Press in The News Media and the Law v35n2 about National Day Laborer Organizing Network v. U.S. Immigration and Customs Enforcement Agency. In this case, a FOIA request was fulfilled with 3,000 pages of documents in unsearchable PDF files. The complaint asks that the government produce the records with "metadata" intact, thus making the files searchable; the response is that in order to redact exempted information properly, it is necessary to produce a new record which obliterates the exempted material completely, and Sharpie plus photocopier does that.
(Well, they didn't say Sharpie, but this is about that level of technology.)
If faced with this situation when you receive documents, what do you do?
1. Appeal the FOIA response, and insist upon searchable documents.
2. Take what you have and apply technology to it, e.g. optical character recognition and other scanning techniques, to create your own searchable index.
3. Take what you have and apply people power to it, dividing up the 3,000 pages into smaller pieces and assigning each part of your national network a small piece of it to review, analyze, and code for later review.
Of course, all three can be taken in parallel.
As an agency, your choices are limited by the systems you have readily at hand. There is redaction software that takes a document in and produces a document in response, with the searchable parts left searchable and the redacted parts fully redacted. If you have this, you will use it, but presumably you will have this only if the cost per redacted page anticipated for the upcoming budget year demands that you apply technology, rather than Sharpies, to the task.
The unanticipated problem for which there are no easy answers is that you are trying to redact only a portion of a record, and thus you need to create a new record with the redacted material gone but the unredacted material intact. If this were a simple database query, you'd just omit the redactable materials; but when the materials are e.g. electronic mail messages, and your policy is to omit phone numbers, you get data dumps that require that you wield a fine point Sharpie.
