-
how do we get more people to participate in anti-censorship
- obfs4 bridge campaign
- how can we get more snowflakes?
-
what do p2p apps use for nat punching?
- georgetown race guy who sometimes lives in china should know
-
a few years ago there was a blog bridge campaign
-
todo: have a contact address in our setup guide for people to complain to
-
should reach out to pets and foci people and ask to set up default bridges
-
ask existing highly-used bridges to turn their bridges into default bridges
-
is it a problem to have default bridges run in the same AS?
- capacity vs safety -> atm we want capacity
- same applies to exits and bridges "myth" -> capacity is more important
-
what kind of usage info is exposed to people?
- approximate number of users per country
-
tor's throttling not accurate enough
- griffin's bridge reached cap 20 days into the month
- default bridge ops need guidance on that
-
orbot has different default bridges for countries
-
would pay off to not have same bridges in tor browser and orbot
- so they're not all blocked at the same time
-
what if harvard runs a bridge, given that china loves harvard?
- at what granularity would gfw block bridge?
-
snowflake status: having things run in a browser tab isn't relevant anymore
- you need an extension these days
-
once you're connected, snowflake should only say "connected"
- only mention numbers once you have > 0
-
how to get more snowflakes
- blog post
- tp.o website frontpage
- talks
- once it gets more popular, people will do advertisement for us
-
roger's defcon talk is in early august
-
cupcake is 2-3k active users each day
- griffin could shove it into package
- there was only about one user at any time
- griffin gave a lot of talks on it (lightning talks etc., workshops, ...)
- cupcake hasn't even gotten press coverage
-
we shouldn't burn our chance to polish it before it gets super popular
- we should coordinate growth and attention
- more important is 1) making sure that it scales, 2) we know what's going on, and 3) the tool is fun to use -> the rest will follow
-
#1 reason for cupcake uninstalls: crazy bandwidth spikes
- rate limiting may also be important
- client and bridge could rate-limit, so the snowflake proxy doesn't have to
-
instead of "how many people have i helped" inform "how many people have all snowflakes" helped? you're part of the snowflake blizzard.
-
broker could inform snowflake proxies how often they should poll again
- that way we can keep the polling rate of all proxies constant
-
cohosh has script on chinese vps that periodically tries to connect to snowflakes
- proxy-go instances running on bridge were blocked for a while
-
does ooni test where webrtc is available?
- we don't think so
- would be difficult to ship webrtc with ooni
- ooni could however test stun servers (todo)
- stun servers work over udp
- tells you what your own external address is
-
snowflake is working on orbot as a client, so they can use it as a bridge
- part of that is also about being a snowflake
- they're ready to turn it on
-
orbot doesn't have moat, so they're in a bad situation when it comes to getting new bridges (copy&paste is also annoying)
- large binary is also a nuisance (our new pure-go implementation would help with that)
-
our new pure-go webrtc implementation also has a media channel
- pionwebrtc is actively maintained and maintainers are super excited to help us in case of blocking
-
todo: what diagrams/graphs/screenshots do we want roger to present?
- needs to be done by july 25
Comments
Please register or sign in to add a comment.