craig@unicus.UUCP (Craig D. Hubley) (03/22/88)
Mark Gooderum asked me to post this for him, since his Pnews is misbehaving: ------------------------------------------------------------------------ I have gotten several messages about Apple Talk and it looks like it isn't as easy as I thought. Would you please forward this to Comp-Sys-Amiga (I had to go throught HUGE contortions to post my message)? Thank you very much for the replies I have received on AppleTalk. I have found out that some of my original assumptions were wrong but here is what I found. 1.) Hardware-AppleTalk uses RS422 **synchronus** hardware protocol. Inside the Mac there is a special chip (by Zilog) that handles things like reconstructing the clock rate. The other problem is AppleTalk runs at 254K baud. The Amiga's serial port is FAST but not that fast. The best I have heard on the Amiga is about 160K baud and that required taking over the machine. In the normal multitasking environment 60K baud seems the limit. So the hardware interface looks like it would require a special port on a board that goes directly on the bus. 2.) Software-Handleing the packets on AppleTalk is an unsure idea. I have heard everything from easy to hard. One person told me that the UNIX version of AppleTalk was 10000+ lines of code. Another person said that as long as you had a 'real' Rs 422 port to handle the hardware part of the protocol the code was easy (like the IBM AppleTalk card.) Also there was some confusion as to terminology as to AppleTalk -vs- 'Local Talk'. Originally AppleTalk was AppleTalk. Then as Apple started to get into connectivity with VAXes and Ethernet, the scope of AppleTalk expanded significantly. Also there were many complaints about the original AppleTalk connectors coming loose because they didn't lock. So Apple came out with new locking connectors. The new connectors are called the "LocalTalk Cabling System". "AppleTalk" now refers to the whole overall system (mainly the protocols) while "LocalTalk" refers to the physical boxes and cables sold. Finally, it looks like a bigger deal than I orininally thought. Several people suggested I contact Apple and the Apple Developer's Association. I will do this. My orininal motivation was a low cost networking solution for the Amiga especailly one that would allow easy access to PCs and Macs. I still think this could be done for LOTS less than Ethernet. Someone said if I cold sell AppleTalk for the Amiga for under $150 I could get rich. I probably could but the whole of this project is too big for just me. So there are a few solutions. 1.) Give up. (Too easy.) 2.) Pair up with some people and attack the problem as a group. This might work if there are interested people out there. If you are interested let me know. I am okay and getting better on programming. (Actually I am a good programmer, just a relatively new serious Amiga programmer.) The group would definitely require a hardware hacker and preferablly an expirience Mac person. 3.) Wait for someone else to do it. 4.) Persuade Apple to do it. Yes, this is a possibility. They put it out for the IBM and they are trying hard to get AppleTalk accepted as the 'new' office network. 5.) Try something simpler. Maybe a SCSI based network. Small but cheap. The main problem is the difference in SCSI drivers. Well, there is my too many cents worth. -Mark Gooderum University of Kansas MARKV@UKANVAX.BITNET
cmcmanis%pepper@Sun.COM (Chuck McManis) (03/24/88)
In article <2412@unicus.UUCP> (Craig D. Hubley) writes for Mark Gooderum : > One person told me that the UNIX version of AppleTalk was 10000+ lines > of code. All depends on whether your kernel supports multiple protocol stacks. It would not take anywhere near 10000 lines of code to put into the Sun kernel. > Someone said if I cold sell AppleTalk for the Amiga for under $150 I > could get rich. I probably could but the whole of this project is too > big for just me. Talk about mixed messages :-) If you 'probably could' get rich why not go for it? No guts no glory. Actually, you are pretty smart to not go for it because you *couldn't* sell AppleTalk HW and SW for the Amiga for $150. When people tell me "You could get rich by selling hard disks for the Amiga for $300." I tell them I could also get rich by selling the brooklyn bridge to people for $300. They miss the part where you have to pay to get the merchandise in the first place :-). >1.) Give up. (Too easy.) This isn't a solution since the problem is still not solved :-). >2.) Pair up with some people and attack the problem as a group. This > might work if there are interested people out there. If you are > interested let me know. I am okay and getting better on programming. > (Actually I am a good programmer, just a relatively new serious > Amiga programmer.) The group would definitely require a hardware > hacker and preferablly an expirience Mac person. From stuff I have seen at Sun it would take a good protocol person and a good driver person about 6 months to go from idea to product, working full time on it. Make that a year if they can only work nights and weekends. The hardware will take about a month to design/prototype/debug. All it is is a serial chip and some decoding gates. I believe a board in the works at Microbotics uses the same chip so the hardware might be a nit. >3.) Wait for someone else to do it. Could be a long wait, since I think even Apple realizes that AppleTalk doesn't scale well to larger networks. I suspect the protocols will remain the same but they might move to something like cheapernet (50 ohm coax) or ethernet on twisted pairs. With the hardware innovations coming out in serial/network chips there is no reason to be a slave to Apple's penny pinching on peripherals. >4.) Persuade Apple to do it. Yes, this is a possibility. They put > it out for the IBM and they are trying hard to get AppleTalk > accepted as the 'new' office network. HA HA HA HA HA HA HA. I'm sorry I didn't mean to laugh in your face like that, it's just that the IBM market is sooooooo huge that you can't ignore it. Apple has traditionally attempted to ignore anything non-Apple, this became impossible in the case of the PC but will be possible with the Amiga for several years. >5.) Try something simpler. Maybe a SCSI based network. Small but cheap. > The main problem is the difference in SCSI drivers. Probably not a solution, since SCSI cables are pretty unweildly with all those wires. The hardware problem will sort itself out fairly quickly. I suggest talking to anyone in the back of AmigaWorld and asking them about there 'serial port' plans. If they are using the right chip (82530 or some such number) then ask if you can beta the hardware and write some drivers for it. If you give them the drivers they might let you keep the hardware. Then talk to Kinetics or someone who has dealt with the AppleTalk protocol, and get some idea of where to get some information. Buy Inside Mac, Vol 1-3. And then go for it. It really is trading time for money in this case, you can get ethernet now for $X (about 900/node) and something else for $Y (you said < $150 above) and M months (I'm guessing about 12). That I know of, there aren't any implementations of AppleTalk so that time factor should probably be treated as a minimum. Solution 6 would be to take an existing serial interface that could do multidrop (RS422 == multidrop) and write a driver for it such that you could use Ron Minnich's TCP/IP or Matt's DNET. That will probably be the fastest/cheapest solution. Of course you could build an RS-422 adapter for the Amiga like the current MIDI adapters and use that as well. --Chuck McManis uucp: {anywhere}!sun!cmcmanis BIX: cmcmanis ARPAnet: cmcmanis@sun.com These opinions are my own and no one elses, but you knew that didn't you.