3 ms·
> and it's impressive that they're so technically savvy they don't need the CSS as noted elsewhere With the logs you provided, they appear to be coming from wi
by Crosseye_Jack 2y ago
> and it's impressive that they're so technically savvy they don't need the CSS as noted elsewhere
With the logs you provided, they appear to be coming from within iMessage.
So when someone posts a link in iMessage it will fetch the favicon(s) and the html in order to generate a “preview” of the page with the title of the page and use one of the favicons. It doesn’t need to fetch any css files to do this.
Not saying bad actors don’t fetch css either, but the lack of it being fetched doesn’t mean that it’s a bad actor.
As for why CF don’t reverse DNS their IPs stating it’s iCloud private relay, well CF are not Apples only 3rd party egress provider (Akamai are also one that springs to mind). So if the number of providers can change at any time, the best source of information about valid egress providers is from Apple themselves.
But Apple do also publish these changes to geo-location databases for you to query, for example: https://www.ip2location.com/demo/104.28.42.8 https://www.ip2location.com/demo/104.28.42.8 lists it as iCloud Private Relay.
As for “are CloudFlare caching my site when ran through private relay?”, not 100% sure, I’ll have to check my own logs and cba’ed right now, but I don’t think so (it’s been a while since I ran tests on it to see how it behaved to be 100% sure right this minute.
But I think it would be silly of them if they did as they may not be aware of the what to cache and for who. Let’s say they cached /profile without knowing what the server is using to determine who the logged in user is, they may false cache-hit and leak data from a previous request. When they act as your sites CDN you explicitly tell them what to cache on, but when acting as a relay (either for apple or their own warp product) for a site they are not a CDN for they are missing this info, sure they could guess, but why risk being wrong?)
- m3047 2y agoThanks for the explanation.