I watched some YouTube videos recently (particularly about tabletop
games, those are quite fun), on youtube.com itself, and it uses those
annoying infinite scrolls. I didn't pay attention to it before, and
heard that infinite scroll proponents claim that those are not
inherently bad, but only when done wrong, and that they are done right
in the huge JS-heavy services (which I generally avoid, so it was hard
to verify) of a few companies, including Google. But evidently it is
pretty bad on YouTube as well: even once you manage to load those to a
position you are interested in (and if you manage to do so: sometimes
it just ceases to load, hanging with a spinner, and you can only try
to reload the page, starting from the beginning), opening a video in
the same tab and then going back in history leads to the top of the
list, with everything else unloaded.
I guess in case of YouTube they at least vaguely make sense in some
situations, for instance when one is mostly interested in the top few
items of a particular sorting, just browsing/discovering videos;
similar to imgur's scrolling (which gets buggy and unusable once you
scroll rather far, though maybe not for everyone). While in the wild
those infinite scrolls are used with kinds of sorting for which
there's no apparent conditions under which the top matches would be
more significant or interesting. Regular pagination has a similar
caveat (i.e., often used just to reduce the load, even if the sorting
isn't needed or important), but at least you can get to a particular
position quickly (or at all) with it, and it can actually reduce the
load in situations where infinite scrolling won't reduce it compared
to just loading many results at once.
It can be useful and/or interesting to attempt to understand views
that are very different from yours, and which seem strange, without
just attributing them to supposed negative qualities of people holding
them, yet sometimes it is rather hard to do. And I have mostly failed
to do so in this case (only came up with it being fancy, fashionable,
and/or akin to visual effects), but a quick search suggested that it
is recommended because then users spend more time on a website (and it
matches the issues described above, though doesn't look like a
positive thing to me as a user), and that it's somehow better for
mobile users (this I can't verify). But I also see them on websites
that don't aim keeping users online longer (e.g., web interfaces of
systems without advertisement), and don't target mobile users. Though
cargo cult (or fashion following) is certainly popular in the software
world, so I guess it would be fair to assume that at least sometimes
those are the reasons.
============
Other news
============
Summer noise continues: as before, the primary sources are kids, dogs,
and apparently semi-deaf people who like awful music. But the summer
is almost over, so there is a hope that it will get better.
And I continue slowly working on WWWLite: introduced tables with
rowspans and colspans, form submission, tabs, history navigation, and
incremental text search in the past few weekends. There's still a lot
to finish and polish (e.g., text search and tab focus for links only
go forward now, it's assumed that only HTML documents are accessed,
multipart encoding for forms isn't supported, a few things can be
optimised, and there's no persistent history or caching at all), but
making more or less steady progress. Now it approaches 3.3 KLOC
without headers, 3.8 with them. Maybe the first non-prototype version
would still fit into 5 KLOC, though a more realistic target may be to
stay under 10.
