Re: [Linux-decnet-user] Feature request
Brought to you by:
chrissie_c,
ph3-der-loewe
|
From: Patrick C. <pa...@ty...> - 2004-05-21 07:06:16
|
On Thu, May 20, 2004 at 04:17:43PM -0400, Stanley F. Quayle wrote: > Let me explain what I'm doing now, and see if this is possible in > Linux DECnet. > > I have a Linux box (RH 7.2) NFS exporting a chunk of disk to my VMS > systems. The VMS systems do daily backups to the Linux box. Also, I > have a whole chunk of data on the Linux box that the VMS systems > "see" as VMS files, characteristics and all. > > How this works is that, when I create a file from the VMS system, two > files get created on the Linux system. Suppose I create ABC.XYZ;27 > on VMS. On the Linux box, I get: > > abc.xyz;27 and > .$ADF$abc.xyz;27 (this is "hidden" because of the leading dot) > > (There's also a "abc.xyz" hard-linked to the "abc.xyz;27" to support > version numbers, but let's not go there.) By the way -- this is all > handled by the VMS TCP/IP stack. > > Anyway, the .$ADF$ file is 90 bytes long, and holds the VMS > characteristics of the file. > > When I access the file from VMS, it gets reported back as whatever > characteristics it had when I created it. If I delete the .$ADF$ > file, it comes back as Stream-LF. > > Not knowing anything about the underlying DECnet protocol, it would > seem that FAL on the receiving system gets the characteristics on > files being sent. And on files being "pulled", it reports the > characteristics back. > > How hard would it be to modify the FAL on Linux to do something like > this? FAL already does this to a limited extent. if you add the -m switch to the daemon (in /etc/dnetd.conf) it will create a .fal directory and put metadata files in there. It can also, to a limited extent, understand the $ADF$ files that the VMS NFS client puts there if you also use -t. A good combination of options for FAL is "-m -t -ae" which will use .fal files where files are createdbut it, $ADF$ files when created by NFS, and use file extension-based guessing for files with no other information. -- patrick |