Changes between Version 9 and Version 10 of Ticket #1005, comment 23


Ignore:
Timestamp:
06/28/16 22:02:31 (5 years ago)
Author:
simonhf@…

Legend:

Unmodified
Added
Removed
Modified
  • Ticket #1005, comment 23

    v9 v10  
    55When the issue does arise -- even if very infrequently -- then isn't it better to handle it gracefully and in an intention revealing way that is also forgiving to clients, and in a way that clients actually understand?
    66
    7 The purpose of the RFCs is to consider all such situations even with garbage input (see [1]), and to show what the correct response should be. IMO a mature service like Amazon ELB or a mature daemon like nginx should take something like Fuzz Testing in their stride and not give 'garbage out' as you suggest. Rather they should behave in a predictable manner according to the RFCs. Do you think otherwise?
     7The purpose of the RFCs is to consider all such situations even with garbage input (see [1]), and to show what the correct response should be. IMO a mature service like Amazon ELB or a mature daemon like nginx should take something like Fuzz Testing without any problems and not give 'garbage out' as you suggest. Rather they should behave in a predictable manner according to the RFCs. Do you think otherwise?
    88
    99As it is, the Amazon ELB to nginx timeout incompatibility could be used in theory as an attack vector. An attacker could send many intentionally incorrect POSTs knowing that nginx will timeout and cause the Amazon ELB to block for 60 seconds, thus over-whelming the ELB upstream connections and blocking the service for others. So this is the type of situation that would be uncovered via Fuzz Testing. It's just a coincidence that I discovered it in regular production traffic.