I’ve just wasted my lunch hour dialling into the BBS I set up last night. It’s slow, it’s crufty, it needs some imaginative Hayes commands, but it works – I successfully logged on using Hyperterminal’s pisspoor Viewdata emulation:

You logged on at 05:01:51
It’s now 05:12:02 on the 22/09/06
You have spent 00:10:10 minutes online
Thankyou for calling The Rabbit Run!

(Yes, I know I need to set the real-time clock!)

Hyperterminal has no Viewdata graphics to speak of although curiously it’s capable of doing background and double-height. It also confuses £#_ which I’d sorta expected it to not do since that’s a fundamental of decent Viewdata emulation.

So, having established that the SIP route works and using the out-of-the-box Asterisk UPCM and APCM codecs means that V23 is fine, I think we’re in a state that the proof-of-concept has succeeded and we can probably go on to put more boards online. Hurrah!

This is a memo to myself, because I’ve been reminded this is the third time I’ve been on a futile hunt around the house: You have no DB9 thru-connectors. None. Don’t even bother looking. Just give up now.

Thankyou.

<%image(20060921-nfs.jpg|120|86|)%>It works! It bloody works!

After much sweating and grumbling, I’ve managed to get three stations working on the Econet using JGH’s Y2k-patched level 3 fileserver software. This now means that the network works (ish) without any non-original Beeb bits. Just need to source some more socket boxes, some cables and preferably a fully working MFM or RLL hard disk because this old Technomatic jobbie is on its way out and I can’t keep prodding the actuator with a screwdriver every time I need it to start up.

Honestly, when I did “*I AM SYST” and it worked, I had to do a doubletake. I’m surprised that it’s working at all!

In any case, this is enough for me to continue with my experiments into the Evil Beeb Project. The worst bit is trying to remember the relevant commands to navigate the directory tree since it’s sufficiently different from ADFS to be confusing. It’s also left me wondering exactly how feasible getting BBS software online using NFS will be. Still, it’s an experience (photos here)!

Right, well, I’m going to tidy up the garage worktops a bit now it’s all working, and go through some of the other Beeb stuff to see if I can get station 3 going with a 6502 copro.

Hooray! I’ve managed to get the EPROM programmer working again, so now have a pile more ANFS ROMs for the Econet. I’ve also connected up the last few socket boxes and mounted them on a piece of timber so they’re not hanging around the place tangling up. I’ve now discovered the drop-cables I thought would work won’t (hint: MIDI cables don’t work as Econet cables ‘cos they’re only wired up to pins 2, 3 and 4) so I’ve hit another roadblock. I’ve got enough DIN connectors for 5 cables but that’s not enough to get the network working.

End result: station 1 and station 2 are properly jumpered, although they’re not talking to each other.

In any case, while researching the pinouts I found a lovely quote:

“We showed Bill Gates the Econet network and he said ‘What’s a network?'”
   — Hermann Hauser (Acorn co-founder), 1982

Fantastic 🙂

So what’s next? I’ve not replaced the BBC power supply for the fileserver, and I need some cables before I can wire up the clock box and the terminators.

*sigh*

Having some time tonight and some peace, I decided to go back to building the Econet and getting all that working, able assisted by the guidelines on on the Beebmaster site.

First up was the EPROM programmer (not strictly part of the Econet but important nontheless). After trying to program some more 27128’s I shifted the programmer to the BBC Model B sitting on the workbench (previously it was connected to the Master 512). Lo and behold, the damn thing programmed a couple of ROMs fine, so I’m blaming either Master incompatibility in the UVIPROM software or the Master’s user port. Either way, I now have new ADFS ROMs done and a copy of the Lancaster Assembler. Just means that the main Model B machine is running at PAGE=&2100 but we all have our crosses to bear.

Secondly I thought I might get the Econet actually wired up with the socket boxes. Over on the Beeb list it’s been suggested that cat5 UTP is fine for Econet wiring. While wiring it up, I noticed a bit of a nasty ozone-y smell coming from the fileserver. “Not good,” I thought again (getting deja vu?), and tried to do a ctrl-break. Nothing. A power-cycle didn’t even boot the damn thing up so I think it’s probably the PSU; no worries, I’ve got plenty of spare BBC Model B units around but irritating nontheless. Meantime, I need to get two 9VDC power supplies (wallwarts will do fine) for the network terminators.

So, current status: ROMs blown properly, station 254 out of action, cabling pretty much done (but not tested) and I’ve even freed up a small bit of worksurface to properly hold the 6502 copro for the fileserver itself.

I’d better do some tidying before Nicky and her sister get home otherwise I’ll be in trouble, but I feel quite satisfied I’ve done some more on this!

<%image(20060907-arenalogin.gif|120|83|)%>Way back in the days when I was a nipper hacking around the school’s Econet, I co-wrote (with a chap called Ben Horsley) my first multi-user game which ran on networked BBC Micro machines. Called Arena II it involved a small map, some objects you couldn’t really do much with, and a hell of a lot of file-handling. I was quite surprised to find it mentioned on the BBC Micro mailing list, referring to Jonathan Harston’s MDFS website here, dated 31st October 1990!

Why file-handling? The reason for this was that I hadn’t really worked out the client/server relationship involved (not that Econet facilitated it too much either) and so the whole system was quite clumsy. Filehandling on Econet was great – if someone else had a file open, it locked it so you couldn’t open it from another workstation. This led to lots of fun in a multiuser random-access system such as Arena, where a program would open the ‘current locations’ file, update the user’s position and read the current positions of other users, then close the file; the only way of working out if a file was open was trying it and trapping the error condition.

This now meant that we had error trapping going on everywhere and whenever an error was trapped, the program counter was reset with the function stack pointer (so if you were in a function or procedure, you suddenly weren’t). A chap called Andy Dowson came up with the novel idea of using GOTO (which wasn’t subject to the function stack) to solve that one – thus it was a haphazard melange of GOTO statements and error trapping.

Add to that the repeated hammering of files (30 people all using the system meant that it was pot luck if you got a lock on the file, which caused vast amounts of network traffic) and the network would slow to a crawl.

I’m tempted to rewrite it to use a couple of service ROM images as a server (or perhaps requiring a 6502 coprocessor), with 6502 machine-code based clients, but I’m not sure I can remember all the 6502 stuff nor (I am reliably informed) do I have time. Great days though.