โจ New release for 47ng/actions-clever-cloud
๐ฅ Save deployment logs to artifacts
๐ฅ Annotate your deployments
๐ Quiet mode to avoid revealing secrets or PII
https://t.co/P5k3JtBw9y
cc @clever_cloud
Quick little feature for @chiffre_io while on holiday ๐จโ๐ป๐๏ธ
Display the min/max browser version for the current period.
This can be super useful when tweaking bundler options to drop support for old browsers.
Tonight's #uidesign for @chiffre_io:
Detailed analytics tooltips, to show:
- OS/browser version distribution
- Screen width distribution for various device types, including useful stats for responsive design:
- Minimum mobile width
- Maximum desktop width
Following this blog post, I tweaked the analytics engine of @chiffre_io to remove immutability where applicable.
On my website's "all time" (includes two HN front pages):
Before: 1 minute
After: 1 second
๐คฏโก๏ธ๐
Thanks @Swizec!
@bmann@steveklbnf@nnnnicholas Apart from a fully P2P design (which has its drawbacks), hosting costs are unavoidable, whether going self-hosted, cloud or blockchain-based.
We prefer to use an indie PaaS. It's simple. It works.
cc @clever_cloud ๐๐คโค๏ธ
๐ข In 30min, I'll talk about my side project @chiffre_io live on the Internetยฎ
๐ What is end-to-end encryption for analytics
๐ง How it works
๐ต๏ธโโ๏ธ The challenges of building a privacy-first SaaS
โจ Live demo
๐บ Join the livestream here:
https://t.co/fASGVVicto
Join us for this Thursday's Fission Tech Talk, on April 29th. @fortysevenfx will be presenting how he built @chiffre_io, a privacy first #e2ee analytics package.
Register to join us https://t.co/V5xWj1Rw4y
๐ฉ Upcoming @chiffre_io#buildinpublic newletter issue:
๐ฆ Adding a status page (featuring @alisalahio & @RozenMD)
๐ Investigating streams for data fetching
๐ฃ๏ธ Speaking at @FISSIONcodes talks on Thursday
Signup here: https://t.co/iVy2CfqJVk
@tdinh_me@chiffre_io For now you would only get an "activity report" by email showing you a volume of visitors graph over a given period.
For the detailed data you'd have to login: sending clear text data to addresses like Gmail would void all encryption efforts.
@BrendanEich@JackEllis@AdriaanvRossum@ay_meshkov@kfranasz Not PHE, but I had a look at FHE at the time which was largely impractical.
I don't think it would change the problem much though, if the heart of the debate is a problem of trust. At least with client-side decryption, the code that runs can be inspected.
@BrendanEich@JackEllis@AdriaanvRossum@ay_meshkov@kfranasz The volume of data is definitely an upper bound on what can be done, for sure.
Some hacks can be done to do the same `reduce` operation client-side and cache aggregates, but won't scale indefinitely (there is only one machine to compute with, vs infinitely scalable servers).
@JackEllis@BrendanEich@AdriaanvRossum@ay_meshkov@kfranasz Unless doing HTTPS end-to-end with the final server, IP and user-agent are target for logging in middleware services (CloudFlare, firewalls, load balancers etc).
But this approach may be difficult to scale.
@BrendanEich@AdriaanvRossum@JackEllis@ay_meshkov@kfranasz This is a simplified diagram of our architecture:
The Chiffre server acts like a message queue for public-key encrypted data. It's decrypted on the client dashboard accessed by 1p, and analytics are calculated in the client.
@BrendanEich@AdriaanvRossum@JackEllis@ay_meshkov@kfranasz Another potential issue is that 3p solutions will still always have server-side access logs (in addition to the data they provide via a tracker script). They can claim to not keep them, but again it's a trust issue, and a harder one to prove since it happens on the backend.
@BrendanEich@AdriaanvRossum@JackEllis@ay_meshkov@kfranasz The approach we're exploring is 3p via end-to-end encryption, so that only 1p has the analytics data.
It shifts the problem from data collection/ownership to a trust problem, that 3p will do what it says.
And that's a wrap for today! Video up on Monday
Join us next week for @karlicoss, and the week after for @fortysevenfx and @chiffre_io
Sign up to attend --> https://t.co/LU4fmISG3b