I saw this the other day, although haven’t had a chance to write about it.

EtherDFS just needs a packet driver on MS-DOS, and it implements it’s own re-director to communicate with a Linux file-server, using it’s own raw protocol.

It certainly looks cool, and looking at how it works, it should be possible to write other drivers to read/write other filesystems for MS-DOS.  It’d be more interesting (to me anyways) if you can write an INT 14 re-director using a 32bit DOS extender to make things easier regarding filesystem ports.

When I get back home, I’ll have to test this on my retro machine, as the idea of just needing a packet driver + TSR sure sounds like a LOT less memory than the Microsoft re-director.

8 thoughts on “EtherDFS

  1. I’ve recently looked for a solution to transfer file to my PS/2 and settled on the FTP server from mTCP. While it works “the other way” (server on the DOS machine, client on Windows) I prefer not to install server software on my primary computer just for my retro machines to connect. The mTCP FTP server worked very well – it can be configured to give me access to all DOS drives so I can both easily transfer files where I want and backup the whole machine.

  2. I like the idea of the DOS machine being the server too.

    But for the more mainstream retro options, maybe there are some more lightweight than Microsoft’s client? Perhaps there is a difference in memory requirements between MS LAN Manager client and the MS Workgroup Client for DOS? I was just reading a review of peer to peer LANs from 1993, it says Netware Lite only needs 13K for a client-only machine. LANtastic was their editor’s choice but I never tried that myself, and unlike MS and Netware solutions I have no idea if there is any Linux software that can talk that protocol.

    • Lantastic uses a custom version of NetBEUI as protocol. In theory you could implement an usermode service in windows talking this custom version of NetBEUI and interface it with the network card using TUN/TAP for Windows. Then you could use Lantastic to connect to it using the minimal client without TCPIP for a less hungry memory communication.Probably something like this can also be done with EtherDFS.

      • Interesting, thanks! I think the article I was reading showed LANtastic had higher memory usage than Netware Lite though, so I’d be inclined to try Netware Lite on DOS as the client and Mars-NWE (Netware Emulator) on a Linux VM as the server, since I used both of those before and it will be a trip down memory lane. I gather from searching the web that Mars-NWE doesn’t work properly on 64-bit hosts, and I imagine it might be hard to get working on a recent Linux distribution, hence the VM. I may run it on Red Hat Linux 6 for nostalgia too!

  3. Is it possible to run a Netware Lite in DOSBOX _with_ network functionality (without I guess no problem) ? In a virtualbox/vmware player this should be no problem if you know the ethernet card (e.g. NE1000) and you got the corresponding driver….

    • If NW Lite runs under dosbox (have no information)

      Because: SLIP xor PPP+packet drivers are supported and packet driver – odi shim exists

      TL;DR try it in order to get experienced, serious serial networking howto is on dosbox wiki

  4. I’ve started using EtherDFS as a solution for networking my own multinode BBS, having fought with Netware way more than reasonable, only to have application software constantly crash when run from the network share.

    I’ve got EtherDFS running in a dedicated QEMU/KVM instance, sharing a 24GB disk image with 16 more QEMU/KVM instances running PC-DOS 7, and MysticBBS 1.07. I still can’t run Mystic from the network share, but have been able to run the EXE from each VM’s C: drive, using the network share for all the shared data volumes.

    On the frontend side, I’m running RLFOSSIL on each of the VMs, and HAProxy on yet another instance that also serves up fTelnet for users to connect with.
    This makes it easy to have a multi-node setup that is transparent to users (one IP to connect to, HAProxy takes care of connecting users to an unused node, and spreads the load around to all the nodes in a round-robin fashion).

    I’ve even gone to the point of making a VNC concentrator that shows all 16 nodes on one screen for easy monitoring.

    The posts here have inspired and informed me to make creating my own multinode BBS a reality. And the recent EtherDFS post is responsible not only for making the multinode networking a reality, but also the EtherDFS sources provided a quick springboard for making a utility that automatically sets the wattcp IP & Mystic NodeID for each of my nodes based on the ethernet MAC address as part of the boot sequence. This has made it very easy to bring up new nodes quickly, and made it possible to scale to 250 nodes very quickly if demand ever called for it.

    I look forward to setting up my BBS fully in the coming months, and making it public – something I had previously given up on just a few months ago.

    Thank you!

    • Dude that is freaking amazing! I too had thought about something like that, but I had given up with my ‘stop gap’ effort of multinode running on OS/2 once I realized I only had multiple users a hand full of times and that even 4 nodes was overkill, but damn if that setup you have going is awesome!

      I’m not sure if EtherDFS supports record locking, you may be in for some trouble when it comes to shared databases and doors.. Hopefully it all works though!

      Let me know when you go live, I’ll be more than happy to feature it, especially with details on your config for spawning nodes, and haproxy.

Leave a Reply