Eudimorphodon
Veteran Member
This is the way I'd go, if I was going to try something like this. It might necessitate a larger "tweener" machine as I'm not sure if you can buy parallel SCSI adapters for modern machines, but I imagine the parts are a lot easier to obtain. Then there's the possibility that on the tweener machine you might be able to use https://ipoverscsi.sourceforge.net/ (which probably requires an old Linux kernel and hence an older tweener machine) and only have to write custom software for one end.
Hah! I'm impressed you were able to dig up that old fossil; I kind of wanted to say that ip-over-SCSI was a thing that people had (in the weirdest of extremities) done before, but I wasn't willing to dig long enough to find it. If you're willing to put the grunt work in you'll find all sorts of weird schemes for cramming IP onto not-really-made-for-it mediums. For instance, a thing that had a bit of traction for a while was IP over HIPPI, it even made it to a draft RFC. (SGI was playing with it for doing crazy-fast LANs between clustered Origin supercomputers.) IP over UltraSCSI feels like, I dunno, a really redneck version of the same idea?
An interesting wrinkle in the description of this project is it requires SCSI cards that are capable of acting as both an initiator and a target at the same time. That raises an objection to an iSCSI bridge that I hadn't even thought of, because I don't think there's any way you can put an initiator behind one of these bridges; it's kind of a non sequitur if you think about it? I mean, I guess in theory someone could have made a bridge that could be configured to map initiator requests coming from a specific SCSI ID to a pre-configured remote iSCSI host, but unless you could find such an animal I assume you'd be stuck having to have the remote host running the initiator constantly sending polls to the end behind the iSCSI bridge to see if it has anything to say and initiate the data transfer from its end.
Good point. Can you read/write a hard drive on a PS/2 at 10Mb/s? I wonder what a PS/2 could even do with 10Mb/s - I presume it can generate and/or sink test data packets at that rate so people can tell what the performance of a network card is, but perhaps if you want to actually display data on the screen it won't be fast enough to do that?
The fastest way to view a modern web page on a PS/2 would probably be to use a VNC-style screen scraper or an X11 server to control a browser running on a computer actually able to render it. Sure, you'll get a little lag, probably a second or two to refresh the window because you'll be pushing a lot of uncompressed pixel data, but a fast 386 or 486 is going to handle that a lot more gracefully than it would trying to grok HTML5. Or SSL.
(Back in the mid-late 1990's I used to display Netscape Navigator running on a computer across campus on a humble 68030-powered hardware Xterminal over 10base-T networking, and it was, well, fine. And it wouldn't be much worse today, assuming I were connecting to a faster computer running the browser.)