I tried setting my UA to '' once and found, sadly, that it breaks all
   sorts of sites whose frameworks or webmasters were too small minded to
   think about the absence of a user-agent.
   As functionally inspired languages become more common and exhaustible
   case blocks filter in, perhaps this problem will be fixed by accident?
   I love your concept of embedding agreements in UA. :)

   On 4 March 2015 00:45:15 GMT+00:00, rysiek <rysiek@hackerspace.pl>
   wrote:

Dnia wtorek, 3 marca 2015 11:50:07 Cathal Garvey pisze:

     Hold on there. These are two different beasts. DNT is "please don't
     track me" and of course it won't work.

     In fact, it's worse. DNT, if set either way, is another pure bit of
     browser entropy; it actually *assists* certain forms of tracking,
     because it can be expected to remain invariant between visits of a
     given
     browser/user.

     Absolutely. However, I did use to give even more bits of entropy bu
     setting my
     UA String in a particular way:
     [1]http://rys.io/en/56
     Now I just need to start filing lawsuits, I guess. ;)

     This is just one of the things making me think the "web" needs a
     total
     re-boot to redesign for security from the boots-up. Servers
     shouldn't
     require user-agents to know how to treat visitors. Scripting is
     useful
     for a rich experience but should be more sand-boxable (ideally,
     scripts
     can be sandboxed to their position in the DOM tree!) and tightly
     permission'd. Canvas and other elements should behave
     deterministically;
     this should be part of browser test-suites. Browsers should be
     allowed
     cache fonts but not disclose to the server whether they have a font
     in
     their cache or not.

     But look, HTTP/2.0 is comming! Oh, wait:
     [2]https://queue.acm.org/detail.cfm?id=2716278

     DNT was another nail in the coffin. Either a browser can be tracked
     by
     design, or it can't.

     +over9000

   --
   Sent from my Android device with K-9 Mail. Please excuse my brevity.

References

   1. http://rys.io/en/56
   2. https://queue.acm.org/detail.cfm?id=2716278