[ntp:hackers] Weekend error report

Danny Mayer mayer at gis.net
Fri Oct 8 16:56:15 PDT 2004

At 02:37 PM 10/8/2004, David L. Mills wrote:
>I chased this one down. Very simple problem. With OpenSSL configured, 
>broadcasts must have the local destination address, so cannot use the 
>wildcard interface. I thought this was clear from the gitgosh.
>It is in principle possible to work around this issue, since the addresses 
>are not necessary for symmetric key cryptography, only autokey. However, 
>the additional hassle doesn't seem justified and probably does not 
>simplify the I/O code.
>The original code, of course, did what was necessary to avoid the wildcard 
>interface and to suck up the destination address for autokey.


It's nice to find out that I was right about this. The I/O code that I'm 
on banishes wildcard addresses altogether, but there are issues about
deciding whether or not the destination address was a unicasy, broadcast
or multicast address. I need to sort that out still. In addition I needed to
write some code to figure out which socket should receive the multicast
packets. It's all in progress right now but the result will be reliable
address/cast information.


More information about the hackers mailing list