XPost: comp.os.vms   
   From: philip+usenet@paeps.cx   
      
   Rick Jones  wrote:   
   > In comp.protocols.tcp-ip glen herrmannsfeldt  wrote:   
   > > I wouldn't be surprised if some broadcast replies.  How many update the   
   > > cache on broadcast requests not destined for them?   
   >   
   > Unless a gratuitous ARP happens to qualify as a reply (even though it isn't)   
   > I don't think that ARP replies are ever sent as anything other than unicast   
   > frames.   
      
   The RFC allows gratuitous arps to be either "requests" or "replies" (actually,   
   it allows it by not forbidding either, if I remember correctly).  In practice,   
   most implementations send gratuitous arps as "requests".   
      
   I have however encountered embedded devices which send them as "replies" (in   
   fact the product I was working on decided to crash if that happened, which was   
   why I looked at the code in the first place :-)).   
      
   My feeling is that gratuitous arps should be "special cased" regardless of   
   whether they say they are a reply or a request.  It's quite easy to imagine a   
   rogue sender of arp messages filling up a host's arp table by sending a large   
   number of gratuitous messages.   
      
    - Philip   
      
   --   
   Philip Paeps                                    Please don't email any replies   
   philip@paeps.cx                                     I follow the newsgroup.   
      
     New systems generate new problems.   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)