Turning Notes into a PDF Worth Archiving

Plain text has no layout, so the converter has to invent one

It is easy to forget how little a .txt file actually contains. It is a sequence of characters, with some of them marking the end of a line, and that is the entire content. There is no page in it, because there is no notion of a page. There is no font, no size, no margin, no indentation, no bold, no heading. When you look at a text file it appears to have a shape, but the shape belongs to the window you are looking at it in — make the window narrower and the shape changes, because your editor is doing the wrapping, live, and storing none of it.

A PDF is the opposite kind of object. It describes exactly where each piece of text sits on a page of a fixed size, which is why it looks the same everywhere and why it can be printed without surprises. Going from one to the other therefore requires supplying every piece of information the text file does not have. There is no correct answer to what those should be, only a set of defensible choices, and a converter that hides its choices is not being modest — it is preventing you from predicting what you will get.

So here they are, read out of the source rather than guessed. The page is A4, 595.28 by 841.89 points; a point is one seventy-second of an inch, so that is 210 by 297 millimetres to the second decimal, which is the ISO 216 definition. The margin is 56 points on each side, about 19.8 millimetres, leaving a text column 483.28 points or 170.5 millimetres wide. Lines are spaced at 1.42 times the font size, so 17.04 points at the default 12. The first baseline sits 68 points below the top of the page and the last one about 58 points above the bottom, which is close enough to symmetric that nobody notices. Forty-three lines fit at 12 points, 65 at 8, 47 at 11, 37 at 14 and 19 at the maximum of 28.

Helvetica is proportional, and that decides what this tool is good at

Every character in a fixed-width font occupies the same horizontal space, which is the property that makes columns line up without any effort: put a value in the twentieth position of every line and it appears in the same place on all of them. Helvetica is not that kind of font. Measured in the tool's own text column at 12 points, a lower-case i is 2.66 points wide and 181 of them fit on a line; a capital M is 10 points wide and only 48 fit. That is a ratio of nearly four to one between the narrowest and the widest character on the same line.

The practical consequence is unambiguous. Anything whose meaning depends on horizontal position — a table laid out with spaces, a log with aligned timestamps, a code listing with indented blocks, an invoice with a column of amounts on the right — comes out ragged. Not corrupted, not unreadable, just no longer aligned, and once you have seen a page of it you will not want to file it. Prose, on the other hand, is exactly what a proportional font is for: at 12 points a mixed sentence averages 5.58 points a character, so about 86 characters fit on a line and roughly 3 700 on a page, which reads as comfortably as any word processor would set it.

There is a second reason the font matters, and it only shows up when someone opens the file in twenty years. Helvetica is one of the fourteen base fonts that every PDF reader has been required to provide since the format was published, so it is referenced by name and its outlines are not stored in the file. That is why a one-page PDF from this tool is under a kilobyte: it contains your text, a page description and a pointer to a typeface the reader is expected to already own. It renders everywhere today and will almost certainly render everywhere tomorrow — but it is a dependency on something outside the file, and the archival standards care about that. More on which below.

Three things that break, and what to do about each

The tab is the sharp one, so it goes first. The text is written into the PDF using a single-byte encoding that has no slot for a tab character, and the conversion does not skip it or turn it into spaces — it stops. What you see is the tool's generic message saying it could not build the PDF, with no indication of why, and the same happens for a form feed and for the whole range of control characters that a log file or a terminal capture may quietly contain. This is a defect and it is being reported as one. Until it is fixed the workaround costs five seconds: find and replace every tab with two, four or eight spaces in a text editor before pasting. Since the font is proportional, the columns were never going to line up anyway, so nothing is lost by the substitution.

The second is the line that is too long to break. The wrapper looks for spaces, and when it does not find one it gives up and writes the whole token on a single line regardless of how wide that line becomes. A 180-character web address at 12 points measures 1 177.7 points and ends at horizontal position 1 233.7 on a page that is only 595.28 points wide, so more than half a metre of it is outside the paper. It is not clipped with an ellipsis, not wrapped onto the next line and not flagged: it is simply drawn where nobody can see it. Long URLs, cryptographic hashes, encoded blobs and minified code are the usual victims. Break them by hand at a slash or a hyphen before converting, and check the right margin of the finished PDF before you file it.

The third is the character the encoding cannot hold. Everything up to U+00FF passes through, which covers the accented letters of all six languages this site is written in — café, naïve, Peña, informação, Straße, città all arrive intact. Everything above that code point is replaced by a question mark. In practice the casualties are the ones you get from pasting out of a word processor or a chat client: curly quotation marks, apostrophes, em and en dashes, the ellipsis character, and — a detail worth noting on a European site — the euro sign, which sits at U+20AC and does not survive, while the dollar sign and the pound sign, which sit lower, do. Non-Latin scripts are lost entirely. Type your quotes and dashes as plain ASCII, spell out any currency you cannot risk losing, and check the first page.

What is actually inside the PDF

Open the result in a viewer and try to select a sentence. It selects, because the text in this PDF is text — a sequence of characters positioned on the page, not a picture of characters. That single property is the difference between a file you can search, copy from, feed to a screen reader and index, and a scan, which is an image of a document and needs optical character recognition before a computer can do anything with it. If the reason you are converting is that a note needs to survive and stay findable, this is the property that matters most, and it is the one this tool gets right.

What the file does not contain is just as worth knowing. Its metadata dictionary holds the name of the library that produced it and two timestamps — the moment you clicked the button, in universal time. There is no title, no author, no subject, no keywords and no language tag. There is no structure tree either, which is the layer that tells a screen reader which text is a heading and in what order the page should be read, so the document is untagged. And there is no XMP metadata packet. Every one of those is a normal, unremarkable omission for a small utility, and every one of them matters if the word archive is being used in its strict sense.

That strict sense has a name and a set of rules. PDF/A, standardised as ISO 19005, is the format that archives and national libraries actually accept, and the Library of Congress summarises two of its requirements in one line each: all fonts must be embedded and legally embeddable for unlimited, universal rendering, and the use of XMP is mandatory for basic descriptive and identifying metadata. The whole point, as the Library puts it, is that PDF/A is constrained to avoid external dependencies. This tool's output fails on both counts — it points at Helvetica instead of carrying it, and it carries no XMP at all. It is a perfectly good PDF. It is not a PDF/A, and if a court, a records office or a compliance regime has asked you for one, this is not the route.

What you have to add yourself

For everything short of formal archiving — a meeting note, a decision record, a set of terminal output kept as evidence, a piece of writing you want frozen — the file is fit for purpose, provided you put the things it cannot store into the only place it has: the text. Start the document with four lines. What this is. When it happened, written as a full date with a four-digit year so it cannot be misread anywhere. Who wrote it. Where it came from, whether that is a system name, a URL or a person. Those four lines cost nothing and they are the difference between a document and an orphan.

Then rename the download. It arrives as document.pdf, unconditionally, every time, and there is no worse habit for a folder you intend to consult in five years. A name that sorts and reads on its own — the date first in year, month, day order, then a short subject, then the source — makes the archive navigable without any metadata at all, which is convenient given that the file carries none. If you produce several notes, merge them into a single dated PDF rather than accumulating fragments; and if the result is going into a mailbox or a ticket, run it through compression first, though at under a kilobyte a page there is usually nothing to compress.

Finally, know when to stop. This is the right tool for a note, a log, a transcript, a licence, a batch of terminal output, a piece of correspondence you want frozen in a form nobody can edit in passing. It is the wrong tool for anything that needs headings, tables, page numbers, a logo, bold text or a second column — not because it does them badly but because it does not do them at all, and forcing plain text into a shape it does not have produces something worse than either. For those, write in a word processor and export from there. The virtue of this converter is that it makes exactly one promise and keeps it: your characters, on a page, in the order you wrote them, in a file that will still open in twenty years.

Frequently asked questions

It says it could not build the PDF and gives no reason. What is wrong?

In almost every case it is a tab character. The text is written into the PDF with a single-byte encoding that has no position for a tab, and rather than substituting or skipping it the conversion stops and the tool shows its one generic failure message. The same happens for a form feed and for the rest of the control characters, which is why the failure is so common on log files, terminal captures and code pasted from an editor that indents with tabs. The workaround is to open the text in any editor and replace every tab with spaces before pasting — two, four or eight, whichever matches how you were reading it. Nothing is lost by doing so, because the font is proportional and the columns would not have lined up regardless. The message being generic rather than saying which character it choked on is a real shortcoming and it has been passed on.

Where did my euro sign and my curly quotes go?

They were replaced by question marks on the way into the PDF. Before drawing each line the tool strips out every character above U+00FF, and the euro sign sits at U+20AC while curly quotation marks, typographic apostrophes, em dashes, en dashes and the ellipsis character all sit in the same neighbourhood. Latin-1 accented letters are below that boundary and pass through untouched, which is why café, naïve, Peña, informação, Straße and città all survive, and why the dollar sign and the pound sign do too. The substitution is broader than it strictly needs to be — the encoding the PDF actually uses does have room for several of the characters being removed — so this is worth fixing and has been reported. Meanwhile: type straight quotes and hyphens, spell out any currency you cannot afford to lose, and glance at the first page before filing.

Why don't my columns line up any more?

Because Helvetica is a proportional typeface and the alignment you had was an illusion produced by a fixed-width font in your editor. In a fixed-width font every character occupies the same horizontal space, so counting characters is the same as measuring distance and columns made of spaces line up for free. Helvetica gives each character its natural width: in this tool's text column at 12 points, 181 lower-case i characters fit on a line but only 48 capital M characters do. Every line therefore ends up a different physical length even when it has the same number of characters, and nothing you type will fix it because the tool offers no choice of typeface. If the alignment matters, this is the wrong tool for that document — take the text into an editor or a word processor, set it in a fixed-width font there, and export the PDF from that. If the alignment does not matter, remove the padding spaces and let the text be prose.

How many pages will my file come to?

Count the lines and divide. At the default 12 points each page holds 43 lines, so a 2 000-line log becomes 47 pages; at 8 points a page holds 65 lines and the same log becomes 31, a saving of a third. Eleven points gives 47 lines a page, fourteen gives 37, and the maximum of 28 gives 19. Bear in mind that the count is of lines after wrapping, not of lines in your file: any source line longer than about 86 characters at 12 points will occupy two or more, so a file of long paragraphs produces more pages than the raw line count suggests. If you would rather think in characters, a full page at 12 points holds roughly 3 700 of them in ordinary mixed-case prose. The tool reports the real page count next to the download button once it has built the file, so the cheapest way to know is to convert once and look.

Is this a PDF/A? Can I use it for legal or regulatory archiving?

No, and the reason is precise rather than a matter of degree. PDF/A, the ISO 19005 family, exists so that a document can be rendered decades from now without depending on anything outside the file. The Library of Congress states two of its requirements plainly: all fonts must be embedded and legally embeddable for unlimited, universal rendering, and the use of XMP is mandatory for basic descriptive and identifying metadata. This tool's output satisfies neither. It names Helvetica rather than carrying its outlines, and it writes no XMP packet at all. It also produces an untagged document, with no structure tree to tell assistive technology the reading order. For an internal note, a personal record or an email attachment none of that is a problem — the file is a valid PDF that will open in any reader for a very long time. For a court filing, a records office deposit, a regulated retention obligation or a national library submission, run the output through a dedicated PDF/A conversion tool and validate it afterwards, since even software that claims to produce PDF/A frequently produces files that fail validation.