mknod and compress corrupting large files
 

To conserve space when backing up projects on our third-party Oracle-based data interpretation system I create a named pipe, hook it up to compress, and feed the backup to the pipe:
mknod <project_name>.pipe p
cat <project_name>.pipe | compress > <project_name>.dmp.Z&
owbackup <project_name> <project_name>.pipe

This works for our smaller projects and works for huge schemas on our HP-UX systems, but we have problems with a project that creates a 65 Gb backup file. Using the above, we get a 45 Gb file that uncompresses to a 46 Gb file and which the import utility reports an Oracle error indicating the file is probably from an aborted export. I can ftp the 65 Gb file to another server with lotsa space and compress and uncompress it with no ill effects on its usability.
:scratch:
Are there any limits on Red Hat Linux mknod pipes or compress that might be causing this?

Thanks!