Questions

Yes. There are 13 demo databases in the workbench, and everything except keeping your work runs on them without an account. They are real exports with the personal data taken out.

Reading, analysing and correcting do not touch it. One screen writes to a source — the Junk Cleaner, which can drop a table, empty one or delete rows by a rule. It does that only when pressed, only with a licence that permits editing data, never on a demo, and it writes a backup of everything it removes first.

They are counted and named. Every column gets a disposition, and anything unaccounted for blocks the export until somebody answers or accepts it. A column nothing names but the target can still hold goes into the target’s catch-all pocket and is reported as kept rather than lost.

The report is built from the same rule the writer uses to fill that pocket, so “kept” on the screen and “kept” in the database are the same claim.

It can still be read: tables, columns, relationships, row counts, and a report of what nothing claims. What is missing is the preset that would name what those tables mean, and you can name it yourself — table by table, on the graph.

Some engines are recognised and deliberately not mapped, with the reason written down. The engines page lists both.

It is a band, not a number, and deliberately so: the estimate is read off a schema, and a schema does not know that one column has had ten years of free text poured into it. A band leaves room for that; a single figure would promise a precision nobody has earned.

Yes — select a row on the Data Instances tab and press Show this row in the target. It is the migration’s own calculation, stopped before the statement goes out, with every correction applied.

It can read your analysis and suggest corrections, and that is all. Every suggestion waits for you. How that works.

Tell us — that is a defect, not a misunderstanding. These pages read the same sources the screens read, so where they differ one of the two is wrong. Contact.