Coming back to where you stopped reading

You are two-thirds through something long — a spec, a longform article, a documentation page, a chapter of a course. Something interrupts. You save it for tomorrow.

Tomorrow it opens, at the top, and you are looking at a headline you already read. Somewhere below is the paragraph you actually stopped at, and finding it means scrolling and skimming and re-reading bits until something feels familiar.

It is a small irritation and it is the reason a lot of long pages get saved twice and finished never. Reopening the page is not the same as reopening the tab.

Why most tools cannot do this

The obvious implementation is to store how far down you had scrolled and put it back. That turns out to be both hard and wrong.

Hard, because restoring a scroll position means running code inside the page when it reopens, and a browser extension reopening a page from a notification has not been given permission to do that. Getting it would mean asking, at install time, for access to every site you visit — for a scroll position. That is not a reasonable trade, and it is why most tools do not bother.

Wrong, because a pixel offset is a fact about one rendering of a document. It is off at a different window size. It is off again once images finish loading and reflow the article, which they always have by the time you come back a day later. It is completely wrong if the site shipped a redesign in the meantime.

What survives all of that is the text you were reading.

Text fragments, which your browser already does

Browsers support a kind of link that points at a passage rather than a position:

https://example.com/article#:~:text=the%20sentence%20you%20were%20reading

Open that and the browser finds the passage, scrolls to it, and highlights it. You may have seen it without noticing — it is what Google uses when a search result jumps you straight to the relevant sentence.

It is the right mechanism for this because the browser does the work, during a navigation that was happening anyway. Nothing is injected into the page and nothing extra is needed.

How it works in Latr

There is a setting called Where you were, in Settings → Appearance. With it on, saving a reminder also stores the line of text that was at the top of your screen. When the reminder fires and you click it, the page opens at that passage, highlighted.

Some deliberate behaviour worth knowing:

Nothing is stored if you were near the top. Most saves are, and there is no position worth remembering.
If the text is gone, the page opens normally. A site rewrote the article, or it is behind a login now, or it renders differently — you get the page as you would have anyway. The reminder is never worse off for having tried.
It is per way of saving. The toolbar button can capture it while the address-bar keyword does not, alongside the same per-gesture switch for page thumbnails.

Full detail is in the appearance settings, and the engineering side is written up separately if you want to know why it was built this way.

Two things that were harder than expected

Reading these back, they are the kind of detail you only find by using the thing on real pages.

Document order is not reading order. Wikipedia’s table of contents sits before the article in the markup and is always on screen, so an early version anchored every Wikipedia save to the word Contents. Anything sticky or fixed, and anything in a nav, header, footer or sidebar, has to be skipped — what you want is the text you were actually reading, not the first text on the page.

Both ends of the passage must land on a whole word. Cutting a passage at 200 characters once ended it mid-word, and a fragment that ends inside a word never matches — so the link silently did nothing. Silently is the bad part.

Doing it without an extension

If you want the same effect manually, most browsers let you select a passage, right-click, and copy a link that points directly at it. Paste that into wherever you keep the page.

It works, and it is three or four steps rather than one, which means you will do it for the important page and not for the other nine. That gap is the whole argument for automating it — not that the manual version is bad, but that you will not do it often enough for it to matter.

Where it earns its place

Long documentation you are working through a section at a time. Courses that are one enormous page. Legislation and standards, where “somewhere in section 4” is not a location. Books published as a single web page. Anything you read across several sittings.

It is genuinely unnecessary for a two-minute article, and it does nothing for a page you were saving for later rather than resuming. That is fine — it costs nothing when it does not apply, because nothing is stored in the first place.