Who it is for

Capture the wording of a regulation verbatim

Regulations, terms of service and official guidance change without notice. A Markdown copy carrying the address it came from and the date the page declared is a record of what the text said on the day you read it.

When the wording is the fact

A supplier updates its terms, a guidance note is reissued, a portal republishes a provision under new numbering, and the version you advised on no longer exists at that address. A screenshot captures a viewport rather than a document. Printing to PDF brings the cookie banner and drops the table of thresholds. And a page run through a generic converter arrives with its clause numbering flattened into prose.

The details that matter are exactly the ones conversion tends to destroy. Fee tables collapse when one cell holds a list. Footnotes become dead anchors back to a page you have left. Sub-paragraph lettering merges into the paragraph above. The result still reads plausibly, which is worse than an obvious failure, because nothing signals that the structure has moved.

There is also the question of where the page goes on its way to your file. A clipper that sends page content to a server for conversion is sending what you were reading, at the moment you read it, to a third party – and what a lawyer reads is itself information about a client. Clean Clipper uploads nothing it reads and asks for no permission on the sites you read, so it can only read the tab where you clicked, only while you clicked. The file it writes has no vendor attached to it either, which matters again years later when the matter file is still required to open.

What survives the conversion

• Verbatim text – nothing inside the article body is summarised, rewritten or reordered.
• Source URL and the page’s declared publication date in the frontmatter of every note.
• Tables of fees, deadlines and thresholds are serialised by the extension itself: 12 of 15 kept on the technical corpus, against 7 for each engine compared.
• Footnotes and reference links become [^1] definitions at the end of the note rather than dead anchors.
• Numbered and lettered lists keep their nesting, so clause structure is still readable.
• Files stay on your own disk: no upload, no account, no third party anywhere in the chain.
• No permission is requested on the sites you read, so the extension can read one tab, at the moment you click it, and has no way to observe browsing.
• A second capture of the same provision never replaces the first. A name collision takes a numeric suffix, so the earlier wording stays on file to compare against.

---
title: "Regulation (EU) 2016/679: Article 17"
source: "https://eur-lex.europa.eu/eli/reg/2016/679/oj"
date: "2016-04-27"
extraction: "dom"
---

## Article 17 – Right to erasure ('right to be forgotten')

1.  The data subject shall have the right to obtain from the controller the
    erasure of personal data concerning him or her without undue delay where
    one of the following grounds applies:

    (a) the personal data are no longer necessary in relation to the purposes …
    (b) the data subject withdraws consent on which the processing is based …

Setting it up for a matter file

The settings below assume the file will be read by somebody else, years from now, who was not in the room. That is a different requirement from reading it tomorrow.

1. Open Options from the extension icon and set the destination to the folder where the matter’s documents live. Put the matter reference in the subfolder field; changing that one field is how you move to the next matter.
2. Leave the icon click as the preview window. On legal material the two seconds are worth it – clause numbering and fee tables are exactly what conversion damages, and the window is where you see it before the file exists.
3. Set the filename template to {date}-{domain}-{title}. The capture date and the publishing body are what identify a version when the same provision exists at four addresses.
4. Turn on every frontmatter field, extraction included. A note recording that the text came from the rendered page rather than from structured data is the sort of detail that answers a question rather than raising one.
5. Set images to skip. On an official page they are crests, share buttons and diagrams that never had text in them, and a broken image link in a matter file reads as something missing rather than as something never captured.
6. Capture the provision page and the “in force from” or version landing page as two separate clips. They carry different dates, and the version page is often the only place the effective date appears.
7. Re-clip the same page at review dates. The second file gets a numeric suffix, and any diff tool then shows what changed between the version you advised on and the version in force now.

Settings for legal material

These differ from the defaults mainly in caution: verify before filing, record the route, and never let a second capture overwrite the first.

## Schedule 2 – Fees payable

| Application                     | Fee     | Deadline        |
| ------------------------------- | ------- | --------------- |
| Initial registration            | £170[^1]| 28 days         |
| Each additional class           | £50     | with the filing |
| Late renewal                    | £50[^2] | 6 months        |

[^1]: £200 where the application is not filed through the online service.
[^2]: In addition to the renewal fee itself.

Three uses in practice

A supplier’s terms, before and after

You clip a supplier’s terms of service on the day you advise on them. Eleven months later the clause on liability caps reads differently, with no version history published anywhere on the site and no notice sent to customers.

Because the second capture took a numeric suffix rather than overwriting, both files sit in the matter folder and a diff tool prints the changed clause in a form you can attach to a note. Reconstructing that from a screenshot would have meant reading two images side by side and hoping.

A fee schedule that keeps its footnotes

An official fees page carries a three-column table where two of the amounts have footnote markers attached, and the footnote text sits at the foot of the page. The table comes across as a table, and the reference links become [^1] and [^2] definitions collected at the end of the note.

The footnotes are the part generic converters lose, because reference links depend on element ids that a sanitiser strips. Here they are collected before the sanitiser runs – an ordering detail that decides whether the qualifications on a fee survive at all.

Guidance reissued under new numbering

A regulator republishes its guidance with the paragraphs renumbered and two of them merged. You have the earlier capture, so the note to the client can state which paragraph of the old text became which part of the new one, quoting both.

The URL is unchanged and the page shows only its current state, so without a capture there is nothing to compare. The publishing body’s own archive may hold a copy; it may also not, and the point of the file is not having to find out under time pressure.

Against the usual ways of keeping a page

The important line in this table is the last one. This tool is a working record, not evidence, and where the distinction matters it matters a great deal.

When the conversion is not clean

The register or search page returned no article

A document register, a case search or a results list is mostly link labels, and the extension refuses any page where more than about a quarter of the extracted characters sit inside links. Open the document itself and clip that. The refusal is deliberate – a matter file containing three hundred link labels labelled as a provision is worse than an empty one.

The clause numbers have gone

Check whether the numbers exist in the page’s text at all. Where a site marks up ordered lists properly, the nesting and the numbering come across. Where the numbering is drawn by the stylesheet – generated counters rather than characters in the document – there is nothing in the text to capture, and no extension can read a number that only exists as a rendering instruction. The fix is to clip a version of the document that carries real numbering – often the print or “plain text” view.

The official version is a PDF

Then this is not the right tool for it. The extension converts rendered HTML, and a great deal of official material is published as PDF only. Where a provision exists in both forms, clip the HTML and keep the PDF alongside it; where it exists only as a PDF, keep the PDF and clip the landing page so at least the address and the date are on file.

The date field is empty, or shows an amendment date

The extension records the date the page declares in its own markup and nothing else. Many official pages declare none, and the field stays empty rather than being filled with the day you read it. Others declare the date of the last amendment rather than of first publication. That is what the page says about itself, and it is repeated rather than interpreted. The effective date, where a site publishes one, is usually on the version page rather than the text page.

What it does not claim

It is not a certified archive – there is no third-party timestamp and no signature, so the file records what you saved rather than proving it to anyone else. It does not read PDFs, which is how a great deal of official material is published. It does not interpret, extract clauses or track amendments. It converts a page. And on a document register or a search result there is no article body, so it reports “no article” rather than returning a list of links.

Questions