I was reading through the NetSpectre paper at:
  http://misc0110.net/web/files/netspectre.pdf

I haven't finished it all yet... key thing was on page 5, which starts with:
      if (x < bitstream_length)
         if(bitstream[x])
            flag = true
                
As I understand it, they are depending error packets that come back to
measure cache fills from a background of packets that fail tests such as
above.

Two things occur to me:
   1) in a really badly bufferbloated network, the
      super-long queues ought to completely screw things up for them.

   2) fq_codel ought to something significant to the measurements.
      I don't know yet if it might them better or worse.  But,
      it ought to do something.

-- 
]               Never tell me the odds!                 | ipv6 mesh networks [ 
]   Michael Richardson, Sandelman Software Works        | network architect  [ 
]     [email protected]  http://www.sandelman.ca/        |   ruby on rails    [ 
        

Attachment: signature.asc
Description: PGP signature

_______________________________________________
Bloat mailing list
[email protected]
https://lists.bufferbloat.net/listinfo/bloat

Reply via email to