[tor-talk] [liberationtech] Not another Haystack right?

[Mike Perry]

Thus spake nathan at freitas.net (nathan at freitas.net):

> As the primary developer of Orbot as well, I also deal with mobile
> latency issues quite a bit as well, and this has led me to think
> quite a bit about how the real time browse and click model of the
> web is highly incompatible with the realities of Tor.

I am a heavy Tor user myself, using Tor Browser almost exclusively for
my web browsing. One major thing I have noticed is that I no longer
browse the web like a normal person does. Most people I observe tend
to browse depth-first, clicking on links of interest in a single tab,
and making heavy use of the back/forward buttons to navigate their
browsing experience.

I however, browse breadth-first. When investigating a topic, I open a
bunch of links in parallel in new tabs, and then just begin reading
the first one to load. I find that parallelizing this latency makes it
manageable, and I don't find myself spending as much total time
waiting on page-load as depth-first Tor users do.

Does this mean we should consider altering Tor Browser's default
behaviors? Perhaps by default, clicking on a link should open it in a
new tab, instead of replacing the current tab, to encourage
parallelism? Or is this too extreme and likely to confuse normal
humans without helping them any?

Link prefetching is a related idea, but it can be fragile and come
with privacy and linkability risk, once you start prefetching across
multiple arbitrary domains automatically without proper identifier and
cache(!) isolation.

Of course, we could also just continue to focus on optimizations to
improve latency in the network itself, instead.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: <http://lists.torproject.org/pipermail/tor-talk/attachments/20111205/c2a4a8ef/attachment-0001.pgp>