v0.0.153

Technology detection recognises the Klaviyo script loader forms browsers actually run (an unquoted or long-attribute src, a preload link, the inline element.src = loader and a proxied script URL) without the loader scan slowing a page that holds many empty scripts or a tag stuffed with openers. Thanks to William for the Klaviyo loader recall fix (#1457, #1463).

Fixed

• Klaviyo detection again recognises the script loader forms browsers actually execute. It accepts an unquoted src, a src after long attributes, a lazy data-src, a preload link, the inline element.src = "//static.klaviyo.com/..." loader and a collected script URL passed through a proxy. The URL signatures remain anchored, and an anchor link, an escaped code sample or a commented-out script tag does not count as a loaded script. Only a rel=preload or rel=modulepreload link counts as a loader, and the inline-loader check stays inside its own script, so a page of many empty script tags is no slower to scan. Known limits of the loader recall: a page that quotes the markers in a raw <pre> or JSON block, an element.src = "..." assignment on a non-script element and an assignment inside a block comment still count. (#1457)