July uptime: 97.33% (down for 19h 47m)*
Expanded Access
Thanks to the help of a few testers in Iran the Wallpunch system is now available for use throughout the country. Testing showed that while TCP connections to the entry nodes were successful the TLS handshake was being blocked. By adding Iran-specific dummy SNIs to the TLS handshake we were able to get the tunnels working.
Once connected the tunnels suffered from extremely high latency and resulting low bandwidth, mainly due to the geographic/routing distance between Iran and existing Wallpunch clusters. Our new Dubai cluster offers much lower latency and better connections for users in Iran, and so far has been quite stable.
I’ve also been working on making Wallpunch a more accessible choice for privacy-focused users. Anonymous ID accounts are now available as an option on the checkout page for those who don’t wish to link their account to an email address.
Of course, anonymous accounts are not particularly useful without anonymous payment methods, so I’ve been researching options for accepting cryptocurrency payments. That should be available in August.
Stability Improvements
Server stability has been improved by tearing down old code paths that were set up to allow v1.0 users to continue connecting while the beta was in progress. In addition some of the servers have been suffering from memory leaks during periods of high usage. These have been alleviated to some extent but additional long-term memory tracking is needed to ensure they are completely resolved.
On the client side, a new Windows version was rolled out to fix a bug that caused Wallpunch to use the wrong network interface for its tunnels on computers with virtual network adapters.
Some users have reported intermittent slowdowns. Initial testing indicates that the slowdowns are not due to server resource exhaustion, but instead are caused by poor connections between clients and entry nodes. Slowdown are more common when clients are connected to certain types of networks. Further testing is needed and will be a top priority in August.
* The Hong Kong cluster was down for around 12 hours due to technical issues with the exit node VPS provider. This is the first time such an event has occurred, and I learned that I need to have backup exit nodes (from fallback providers) available to be swapped in on demand.