Better security and privacy through image proxying(blog.fastmail.fm) |
Better security and privacy through image proxying(blog.fastmail.fm) |
Unless, of course, the image has an email-unique key in the URL somewhere.
Overall though, its still an improvement. A key in the URL does verify that the email address is deliverable, which isn't nothing, but still can't be directly correlated with an IP address or other tracking data.
(No c99.php.gif for you!)
Once we decided we wanted to do this, then we looked around to see what other places were doing. That's when we noticed Gmail, but it was actually Github's Camo proxy that we thought of first (and ended up cloning).
The content of the blog post strongly suggests that increased privacy was the primary reason and that the padlock change was secondary.
I'm surprised you weren't aware of GMail rolling this feature out. It was big news at the time. Not least because a lot of advertising companies were worried it would affect their bottom line.
It was the initial motivator. The privacy advantages however are still real. Ultimately it all goes to our customers being able to have confidence that when they're using the service their security and privacy are being taken care of.
> I'm surprised you weren't aware of GMail rolling this feature out. It was big news at the time.
Honestly, I can't remember if I noticed when GMail rolled it out. Looking back on the press from the time, I imagine that if I thought anything it was "oh, you too" because I strongly remember Github rolling out a similar feature in 2010.
https://github.com/blog/743-sidejack-prevention-phase-3-ssl-...
And sure, its not exactly the same thing, but close enough that it might not have stood out to me.
Did others at FastMail notice? Maybe. I certainly don't have a record of a conversation about it, and the first discussion we had about implementing our own image proxy was in May, and Gmail was not mentioned until much later once we decided what we wanted to do and starting shopping around for an off-the-shelf package.
I'm not sure what point you're trying to make. Or are these just observations?