I find that a bit surprising as well. Though my threat model doesn't currently account for state actors, if it did, this would be concerning.
I find that a bit surprising as well. Though my threat model doesn't currently account for state actors, if it did, this would be concerning.
Further if you are a US Citizen talking with Other US Citizens, then having your data in the US is the best place for it... Putting it outside the US automatically eliminates any of the paper thin constitutional concerns the state actors would have to pretend to care about... If your email in a 5 eyes nation then it makes their job even easier.
And regarding the privacy risk, we haven't seen a credible threat that would have our servers accessed directly by US law enforcement that isn't just as credible anywhere else in the world. It hasn't happened, we don't believe it will happen.
So yeah, we disappear in an interesting little thought experiment like this one (despite having Australian-owned IP space, it's advertised and routed into our US locations, so fair enough it's blocked in this case)
I do think the timing is ripe for us to look at having more datacenters around the world again. We have had both Iceland and Amsterdam locations in the past, and neither were reliable enough to put production traffic to, multiple hour network outages and inability to contact support and get them fixed - but maybe the landscape has changed enough that it's viable again now.
Fastmail's clients are mostly online only and dont cache much information.
It takes ~5-10 seconds to go from clicking on a notification to actually being able to read anything. It's really not a good experience.
https://paste.debian.net/plain/1282885 (Zoom out a bit to see it properly)
I also created a few measurements on RIPE Atlas.
Ping from all probes in India to api.fastmail.com: https://atlas.ripe.net/measurements/55469007/
Traceroute from all probes in India to api.fastmail.com: https://atlas.ripe.net/measurements/55469042/
It's not massively worse than what I get from here:
11. 103.168.172.63 0.0% 5 251.0 251.2 250.8 251.9 0.5
We're using cloudflare for DDoS protection and hence our packets all go into their edge where they get scrubbed before being passed to our hosts. They don't inspect the content at all, so they can't terminate early, which means every packet is round tripping that 350ms.
BUT - JMAP should still mean it's only a couple of round trips to have a fully loaded mailbox.