On Monday 03 February 2003 20:09, Peter W Schmiedeskamp wrote: [...]

Our router equipment allows for multicast, however the broadcast discovery packets are killed. So, it seems that I'm not able to image a machine on say 192.168.6.X subnet from 192.168.7.X subnet.

Am I just missing an option, or is it possible to specify an IP address that the clients can listen to, or possibly a way to specify a list of IPs to include in the multicast offer?

I've tried the TTL option, but it is at the router level that a broadcast offer is killed.

-Peter

Unfortunately, without full multicast support from the router, it's not possible to use multicast accross subnets.

There are possibilities to use a different multicast address for the discovery packet (-M option, to be given both to the sender and the receiver), just in case the router selectively blocked address 224.0.0.1 but let pass other addresses.

However, support for this alternative is incomplete right now: - on the server, you also need to supply --async and --max-bitrate options. Preferably --fec too. - receivers must be ready before sender.

./udp-sender --max-bitrate 10m --ttl 2 -f /dev/zero -M 224.1.2.3 --fec 8x8 ./udp-receiver --ttl 2 -f /dev/zero -M 224.1.2.3

[...]

I try to build the software on FreeBSD, but it fails. It seems that udpcast uses getopt.h instead of unistd.h for handling the command line options.

Maybe you could try just replacing every occurrence of #include <getopt.h> with #include <unistd.h> You might also need to replace the calls to getopt_long by calls to getopt. This would make all options that have no one-letter shorthand inaccessible, but those options are rarely used anyways, thus this would not be such a big issue.

[...]

Was there much magic in porting to Cygwin? Cygwin seems to try to duplicate the Linux included headers like getopt.h.

Most Cygwin changes are due to a slightly different network API, and due to a less precise timer granularity.

[...]

Thanks a million. I've also got a tip for creating compressed system images:

Zero out the harddrive before installing the OS and other software on the machine from which the snapshot will be taken. This will make sure the disk is blank. When you are compressing the image, those zeros scrunch down pretty small! IBM has a tool for zeroing the harddisk called wipe(http://service.boulder.ibm.com/storage/hddtech/wipe.exe). Otherwise, booting from a Linux / Unix floppy and doing a 'dd if=/dev/zero of=/dev/hdX' will do it too.

This wipe tool is indeed very interesting, especially if your Windows is on an NTFS partition, which is not as easy to access from linux. Thanks for the pointer.

Regards,

Alain