I tend to keep the line I'm reading at the top of the page. It's annoying if I want to reread something because I missed comprehension, but it makes me faster than needing to figure out where the next line begins while performing a visual carriage return and line feed.
Personally, I love infinite scrolling when I'm consuming content. The problem I face is when it is online content and I've followed a link. This isn't a problem for websites I view in a computer, because I usually just open a new tab and when I close it, I'm right back where I left off. If the site or application stopped after only 20 items, perhaps it was that 21st item I would have really been interested in reading.
For the past several years I've mostly switched to devices like my phone and tablets. Applications don't often feature the same approach and can't as easily return to the view I was looking at. This really limits the usefulness of infinite scrolling. If the application unloaded in the background, when I return to it, the view has reset itself at the top of the list. If I'm a hundred "pages" into the stream, I need to start over with new content that has arrived since I started reading, scroll for a few minutes until I start to recognize content that was close to where I left off, and then actually find where I went off on a tangent.
One solution to fix this problem would be a user enabled filter that hides content that I've scrolled past and only shows "fresh" content. Pagination usually solves this problem for you by having a view that includes where you are in the list. Of course that is invalidated when the whole stream is reset, but if only 10 new items have been added, it wouldn't take that long to resume where I left off.
Finally, sites that weight the ranking, like Reddit or Hacker News, cause both systems to foul up. Since the content resorts it is practically impossible to determine "where you left off." Implementing a mechanism to hide headlines for posts I've already scrolled past, would go a long way to resolving this last problem. In a best case scenario, I'd be more productive. In a worst case scenario, I'd be no better off than I am today, but it is difficult to think of situations where this design wouldn't be beneficial to the end user.
Personally, I love infinite scrolling when I'm consuming content. The problem I face is when it is online content and I've followed a link. This isn't a problem for websites I view in a computer, because I usually just open a new tab and when I close it, I'm right back where I left off. If the site or application stopped after only 20 items, perhaps it was that 21st item I would have really been interested in reading.
For the past several years I've mostly switched to devices like my phone and tablets. Applications don't often feature the same approach and can't as easily return to the view I was looking at. This really limits the usefulness of infinite scrolling. If the application unloaded in the background, when I return to it, the view has reset itself at the top of the list. If I'm a hundred "pages" into the stream, I need to start over with new content that has arrived since I started reading, scroll for a few minutes until I start to recognize content that was close to where I left off, and then actually find where I went off on a tangent.
One solution to fix this problem would be a user enabled filter that hides content that I've scrolled past and only shows "fresh" content. Pagination usually solves this problem for you by having a view that includes where you are in the list. Of course that is invalidated when the whole stream is reset, but if only 10 new items have been added, it wouldn't take that long to resume where I left off.
Finally, sites that weight the ranking, like Reddit or Hacker News, cause both systems to foul up. Since the content resorts it is practically impossible to determine "where you left off." Implementing a mechanism to hide headlines for posts I've already scrolled past, would go a long way to resolving this last problem. In a best case scenario, I'd be more productive. In a worst case scenario, I'd be no better off than I am today, but it is difficult to think of situations where this design wouldn't be beneficial to the end user.