Scrolls were abandoned because they became unwieldy much faster than a codex. Essentially, they were much more work for the user (in terms of reading, holding and storage). The opposite is true of text on the Web — navigating between pages is more trouble than scrolling downward. In the worst case, scrolling down a page in a digital document is as much work as navigating to a different page, and there are a lot of ways it can be easier. One long document doesn't necessarily take up much more physical space or cognitive resources than a short one, but a whole bunch of related documents are often harder to navigate than a single one.
Remember, scrolling was the dominant way of reading documents on a computer long before the Web came along. Some programs used a page interface, but again, it was generally more trouble, and scrolling won out on its merits. The only areas where paging dominates over scrolling are those meant to emulate physical media (e.g. ebook readers) and those where reader satisfaction is a secondary goal (e.g. news sites that divide stories up to boost ad views).
