[comp.sys.amiga] AppleTalk, Networking, Amigas and Offices

zofka@cernvax.UUCP (zofka) (04/15/88)

And now, what about networking with the Amiga.

There is a big hole in the SoftWare market. Amigas now need Networks.

As we all know, every machines present in offices use networks.
AppleTalk for Mac and few others for IBM. Amiga has no capability
to communicate with others ones (except with DNET (1 other)).

I can't recommend Amiga to peoples how need it in offices. Ethernet is
to expensive and not easy to install.

Last month, I saw an article about AppleTalk on comp.sys.amiga.

The author said that he can write the code but the amiga has no capability
to send packets at high speed thru serial port...

If you need an cheap, low-cost, auto-config, zorro,... board for your
software, I CAN DO IT. I have currently two quiet terminated development
and it's very easy to implement fast serial lines (specialized chips
exists and doesn't cost a lot). 25 chips on a card, including auto-conf
and buffering and for 200$ we can have an AppleTalk... And for A500
thru A99999999 ;-).

I have no time to develop software like networks but a small card is only
few weeks to develop (I can write the driver too if you need).

Time to create the NDG no ? (Network Development Group)...

Interested ??? Just reply on the Net or by mail

zofka@cernvax.UUCB
zofka@cernvm  (Bitnet)
milan@BIX     (Bix)

Milan Zofka 26, rte d'Annecy  CH-1256 TROINEX  GENEVA SWITZERLAND
=========================================================================

Last thing.... Excuse my so bad english... Do you prefer well written french?

root@sbcs.sunysb.edu (SBCS Systems Staff) (04/17/88)

In article <648@cernvax.UUCP>, zofka@cernvax.UUCP (zofka) writes:
> And now, what about networking with the Amiga.
> 
> There is a big hole in the SoftWare market. Amigas now need Networks.

	Amiga has had networks for at least 1-2 years.  Ameristar supplies
	IP/TCP, NFS, Telnet, etc and an Ethernet board for 2000's (500's
	soon).  Catina supplies a peer to peer network.  I think Nine Tiles
	in England supplies some sort of token ring board.  Although I
	personally am hard pressed to call it a network, CLtd supplies
	some software for their SCSI controllers which allows disk sharing.
	I heard people talking about TOPS, Appletalk, DECNET, ISO,
	etc, etc.  There is Dillon DNET.  UDEL has a PD port of Phil Karns
	excellent IP/TCP package.  Dale Luck is working on Xwindows window
	server.  Ameristar is working on Sun's NeWS window server.

	YES zofka, the Amiga has networking.

> 
> As we all know, every machines present in offices use networks.
> AppleTalk for Mac and few others for IBM. Amiga has no capability
> to communicate with others ones (except with DNET (1 other)).
> 

	You're wrong about the Amiga talking to other Amigas.  Also,
	in the States, Ethernet is probably the single most important
	network at least in the scientific/engineering markets.

> I can't recommend Amiga to peoples how need it in offices. Ethernet is
> to expensive and not easy to install.
> 
	Ethernet costs what it does because of the volume that the
	Amiga market can support.  For example, Ameristar would do
	cartwheels and drop their prices if they could ship 10% of
	what 3Com does in the PC market.  It takes a lot more money
	to run a company + pay even small salaries (& overhead), power
	bills, phone bills, etc than you realize.  All of this gets
	amortized into the board cost as overhead and of course the
	more boards you ship, the less overhead you pay/board.

	I would also hasten to point out that, for example, Suns PC/NFS
	plus a 3Com PC Ethernet board is $900 last time I checked.  This
	is what Ameristar charges for equivalent functionality on the
	Amiga.  And I can tell you that they sell at least 10X as many
	packages as Ameristar does (PC/NFS supposedly sold 2000 units on
	the first day orders were accepted).

	If you're complaining that Ethernet is inherently more 
	expensive, well I think you're wrong but I will argue the 
	point via e-mail.

> Last month, I saw an article about AppleTalk on comp.sys.amiga.
> 
> The author said that he can write the code but the amiga has no capability
> to send packets at high speed thru serial port...
> 

	Not quite.  The Amiga hardware is not compatible with the hardware
	attributes of the Appletalk net.  It's more than just a level
	shifting problem too!

> If you need an cheap, low-cost, auto-config, zorro,... board for your
> software, I CAN DO IT. I have currently two quiet terminated development
> and it's very easy to implement fast serial lines (specialized chips
> exists and doesn't cost a lot). 25 chips on a card, including auto-conf
> and buffering and for 200$ we can have an AppleTalk... And for A500
> thru A99999999 ;-).
> 

	I've looked at it, and I can tell you that you will not be
	able to build it, sell it, and make money at a $200 price point.
	In the first place, you will not be able to play the same
	games as Apple did in their implementation, ie basically have
	the processor poll the serial port.  The data comes in at
	about 230.4 Khz/8, and you can't catch it byte by byte with the
	processor and still multitask.  You will need DMA or at least
	a small processor + buffer ram on the card.  All of this drives
	the cost up.  

	The other problem with Appletalk, at least if you listen to people
	who have them running is that it doesn't scale, it doesn't give
	you the bandwidth you need, etc.  We ought to learn this lesson
	from the Mac people before we go out and build a network interface
	that is already obsolete.

> I have no time to develop software like networks but a small card is only
> few weeks to develop (I can write the driver too if you need).
> 

	If you have no time to develop, then you clearly have no time to
	do the support work.  Ameristar spends quite a bit of telephone
	time doing support, eg helping people debug their networks they
	already have installed before the Amiga ever arrived!

> Time to create the NDG no ? (Network Development Group)...

	I don't think so, but then I am biased. 

> 
> Interested ??? Just reply on the Net or by mail
> 
> zofka@cernvax.UUCB
> zofka@cernvm  (Bitnet)
> milan@BIX     (Bix)
> 
> Milan Zofka 26, rte d'Annecy  CH-1256 TROINEX  GENEVA SWITZERLAND
> =========================================================================

						Rick Spanbauer
						SUNY/Stony Brook
						(& Ameristar)

dillon@CORY.BERKELEY.EDU (Matt Dillon) (04/18/88)

:The author said that he can write the code but the amiga has no capability
:to send packets at high speed thru serial port...
:
:If you need an cheap, low-cost, auto-config, zorro,... board for your
:software, I CAN DO IT. I have currently two quiet terminated development
:and it's very easy to implement fast serial lines (specialized chips
:exists and doesn't cost a lot). 25 chips on a card, including auto-conf
:and buffering and for 200$ we can have an AppleTalk... And for A500
:thru A99999999 ;-).

	It's actually easier than that.  All you really need to do 
is get some open collector drivers (IEEE-488 style) and connect them to
the serial port.  Add a couple of resistors and Poof, ethernet style network.
With a custom serial driver you could go quite fast.  All that's left is to
write a protocol for it.  Unfortunetly, this is no mean feat.

				-Matt

zofka@cernvax.UUCP (zofka) (04/19/88)

Newsgroups: comp.sys.amiga
Subject: Re: AppleTalk, Networking, Amigas and Offices
Summary: 
Expires: 
References: <648@cernvax.UUCP> <1151@sbcs.sunysb.edu>
Sender: 
Reply-To: zofka@cernvax.UUCP ()
Followup-To: 
Distribution: 
Organization: CERN European Laboratory for Particle Physics, CH-1211 Geneva, Switzerland
Keywords: AppleTalk, Network

>> If you need an cheap, low-cost, auto-config, zorro,... board for your
>> software, I CAN DO IT. I have currently two quiet terminated development
>> and it's very easy to implement fast serial lines (specialized chips
>> exists and doesn't cost a lot). 25 chips on a card, including auto-conf
>> and buffering and for 200$ we can have an AppleTalk... And for A500
>> thru A99999999 ;-).
>> 
>
>	I've looked at it, and I can tell you that you will not be
>	able to build it, sell it, and make money at a $200 price point.
>	In the first place, you will not be able to play the same
>	games as Apple did in their implementation, ie basically have
>	the processor poll the serial port.  The data comes in at
>	about 230.4 Khz/8, and you can't catch it byte by byte with the
>	processor and still multitask.  You will need DMA or at least
>	a small processor + buffer ram on the card.  All of this drives
>	the cost up.  

   I'm not so sure... A processor costs 2-3 $ 32k memory 2$ one pal
   for 1.5 $ and logic for 10 $. The printed circuit is the most expensive
   and the connectors (50 $)... all for 100$. Why every time people talk
   about profit. Few $$$ for few hours of work is enough for people how
   love their work...
>
>	The other problem with Appletalk, at least if you listen to people
>	who have them running is that it doesn't scale, it doesn't give
>	you the bandwidth you need, etc.  We ought to learn this lesson
>	from the Mac people before we go out and build a network interface
>	that is already obsolete.

   Obsolete but they sell it every day.. I.B.M is also obsolete. New
   machines are also obsolete but they sell thousand and thousand of
   pc because there is a lot of software, lowcost networks a.s.o.
   (& nice screens (personnal opinion!!))

>> I have no time to develop software like networks but a small card is only
>> few weeks to develop (I can write the driver too if you need).
>> 
>
>	If you have no time to develop, then you clearly have no time to
>	do the support work.  Ameristar spends quite a bit of telephone
>	time doing support, eg helping people debug their networks they
>	already have installed before the Amiga ever arrived!
>
>> Time to create the NDG no ? (Network Development Group)...

   We are talking about a development group how don't look for real
   profit and $$ but that can help commodore and Amiga to be at the top.
   Imagine, we have a multi-standards-compatibility system... The ameristar
   Ethernet card is a great example but I have a friend with a little
   store and 2 amigas for bills and stock control... Ethernet and
   2000$ investment ??? Impossible 500$ yes...

   zofka@cernvax.UUCB
   zofka@cernvm  (Bitnet)
   milan@BIX     (Bix)

   Milan Zofka 26, rte d'Annecy  CH-1256 TROINEX  GENEVA SWITZERLAND
   =========================================================================
>
>						Rick Spanbauer
>						SUNY/Stony Brook
>						(& Ameristar)

   Ohhh, one more thing, People talk a lot about products for the amiga
   and networking. The only two things I have seen up today are your
   ethernet product (nice) and DNET (nice too but I don't have it!).
   The others ??? waporware ? I have a CLtd Disk and I wrote last year
   two times to CLtd to have informations about SCSI network and
   opticaldisk.device (SCSI too)... Nothing... No reply, no answer...
   nothing...

zofka@cernvax.UUCP (zofka) (04/19/88)

In article <8804180638.AA23679@cory.Berkeley.EDU> dillon@CORY.BERKELEY.EDU (Matt Dillon) writes:
>:The author said that he can write the code but the amiga has no capability
>:to send packets at high speed thru serial port...
>:
>:If you need an cheap, low-cost, auto-config, zorro,... board for your
>:software, I CAN DO IT. I have currently two quiet terminated development
>	It's actually easier than that.  All you really need to do 
>is get some open collector drivers (IEEE-488 style) and connect them to
>				-Matt
   Not so easy at all, look at max speed of the amiga (192 kbits) and
   this with 100% cpu time... 230kbits for appletalk and with alittle
   card, something like 10% cpu (or less) Few more $$ but more efficient.
				 Milan (zofka)

root@sbcs.sunysb.edu (SBCS Systems Staff) (04/20/88)

In article <651@cernvax.UUCP>, zofka@cernvax.UUCP (zofka) writes:
> Newsgroups: comp.sys.amiga
> Subject: Re: AppleTalk, Networking, Amigas and Offices
> Organization: CERN European Laboratory for Particle Physics, CH-1211 Geneva, Switzerland
> Keywords: AppleTalk, Network
> 
>    I'm not so sure... A processor costs 2-3 $ 32k memory 2$ one pal
>    for 1.5 $ and logic for 10 $. The printed circuit is the most expensive
>    and the connectors (50 $)... all for 100$. Why every time people talk
>    about profit. Few $$$ for few hours of work is enough for people how
>    love their work...

	Milan, I love my work.  Couldn't really think of anything better
	to do.  Need to make bucks (== profit) to pay rent, buy food for 
	my wife and child, pay car insurance, etc.  I'm sure the people 
	who provide the services that I've mentioned probably love their 
	work too, but they're in the same boat as I am about paying for
	their necessities.  Right down to the guys who grows the food, 
	dig iron ore, etc.  The basic theme here is $$$ = goods/services.

	As I've pointed out to you in private mail, parts cost are
	a factor in determining price, but are NOT the only force.  Big
	companies charge at least 3X-5X over manufacture cost in order to
	remain profitable.  But I'll promise you this:  if everyone else works
	for free, and doesn't charge for services, then I will too.  I suspect
	that to arrive at this "ideal" I'll have to move east to a colder
	climate and learn to like vodka and copies of 8080's :-).

> 
>    We are talking about a development group how don't look for real
>    profit and $$ but that can help commodore and Amiga to be at the top.
>    Imagine, we have a multi-standards-compatibility system... The ameristar
>    Ethernet card is a great example but I have a friend with a little
>    store and 2 amigas for bills and stock control... Ethernet and
>    2000$ investment ??? Impossible 500$ yes...
> 

	Gee, Commodore US is operating in the "black", financially speaking.
	Is this not true of Commodore/France?  CA seems really very capable
	indeed of helping themselves.  Naturally anyone selling Amiga based
	products would want CA to succeed.  The way we support their
	success is by trying to bring to market the sorts of products users
	would want to buy.  Not by contributing free labour.

	A development group is an OK idea for research sorts of things or
	where ideas need proving (eg there are quite a few superb groups
	in the Amateur Radio area: ka9q ip/tcp folks, TAPR, AMRAD, etc).
	To answer the telephone and provide support for your customers
	pretty well requires that you charge enough $$$ to pay someone to
	spend 40 hrs a week waiting for calls.  

>    zofka@cernvax.UUCB
>    zofka@cernvm  (Bitnet)
>    milan@BIX     (Bix)
> 
>    Milan Zofka 26, rte d'Annecy  CH-1256 TROINEX  GENEVA SWITZERLAND
>    =========================================================================
> 

[stuff about Ameristar & Dillon DNET deleted]

>    The others ??? waporware ? I have a CLtd Disk and I wrote last year
>    two times to CLtd to have informations about SCSI network and
>    opticaldisk.device (SCSI too)... Nothing... No reply, no answer...
>    nothing...

	I believe a few folks at Westchester have the CLtd disk sharing
	stuff running on Amigas.  I know that a few of our customers have it 
	also.  Perhaps the literature that CLtd shipped to you fell down
	the same blackhole that our literature did?

						Rick Spanbauer
						(Ameristar)
						(516) 698-0834

sdean1@uvicctr.UUCP (Steven A. Dean) (04/24/88)

In article <655@cernvax.UUCP> zofka@cernvax.UUCP () writes:
[stuff deleted...]
>>	It's actually easier than that.  All you really need to do 
>>is get some open collector drivers (IEEE-488 style) and connect them to
>>				-Matt
>   Not so easy at all, look at max speed of the amiga (192 kbits) and
>   this with 100% cpu time... 230kbits for appletalk and with alittle
>   card, something like 10% cpu (or less) Few more $$ but more efficient.
>				 Milan (zofka)

Well, let's see here... 192 KBits you say?  Hmm... that's not very much now...
if I do my math here I get some interesting figures... 

      Processor speed  :    7.16 MHz.
      Mem Wait States  :       0 (for a workbench screen)
      Time for a move  :      12 cycles (move.b  (a0),(a1)+)
      Time for a dbra  :      10 cycles (dbra    d7,loop)
      Total CPU time   :      22 cycles per byte.

      7,160,000 / 22 = 325,413 BYTES per second.  

Well, we all know that those speeds are pretty hairy and would require the
entire system to go dead... but I can't see it being too difficult to get about
100K per second... A friend of mine got (with some UGLY code) about 500KBaud.
The serial port is pretty quick...  
 
BTW, does anyone know how fast data can be pushed through the parallel port
with any reasonable amount of efficiency?  I've never played with it... It 
might be an idea to see how well it can grab data.

Steve
--
Steven A. Dean
sdean1@uvunix.BITNET
sdean1@uvunix.UVic.CDN
...uw-beaver!uvicctr!sdean1
...ubc-vision!uvicctr!sdean1

snyderw@pawl17.pawl.rpi.edu (Wilson P. Snyder II) (04/25/88)

In article <392@uvicctr.UUCP> sdean1@uvicctr.UUCP (Steven A. Dean) writes:
>100K per second... A friend of mine got (with some UGLY code) about 500KBaud.
>The serial port is pretty quick...  
> 

Now, is there ANY way to use the BLITTER to get/put data from one address
into/outof memory.  If the system could DMA the data you could have a very
fast parallel port (assuming that you got the data in the short period
before the next byte was written.)  I assume that one problem would be the
need for chip memory.  Comments?





______________________________________________________________________
Wilson P. Snyder II		snyderw@pawl.rpi.EDU
318 Crockett Hall, RPI		518-276-2764
Troy, NY 12180-3590		802-658-3799 in summer
______________________________________________________________________

dillon@CORY.BERKELEY.EDU (Matt Dillon) (04/26/88)

>Now, is there ANY way to use the BLITTER to get/put data from one address
>into/outof memory.  If the system could DMA the data you could have a very
>fast parallel port (assuming that you got the data in the short period
>before the next byte was written.)  I assume that one problem would be the
>need for chip memory.  Comments?

	The question is, how to create a local area net without buying
(or designing) a real-life DMA driven auto-configuring board.  The problem
is really how to deal with multitasking.  One could write a fully 
handshaking scheme driven by a task over the parallel port that would
have an aggregate throughput on the order of, say, 60 to 100 KBytes/sec.
Over the serial port, a fully handshaked scheme could run at about 
26KBytes/sec.  Using an audio channel, one could easily output a 16 symbol
set at 28KBaud = 14KBytes/sec (assuming the filter is turned off), though 
getting data back would pose a problem.

	Now, one must deal with multiple machines on the network.  One
cannot simply connect multiple TTL/CMOS outputs together.  Usually something
like an open collector buffer is used (wire-and).  The serial line has
the fewest number of wires ... two, thus two o.c. drivers.  The network
itself would be a single data wire and a single ground reference.  
Alternately one might want to capacitively couple the network instead of
using O.C. drivers....  now if only I had some time...

				-Matt

mph@rover.UUCP (Mark Huth) (04/30/88)

In article <392@uvicctr.UUCP> sdean1@uvicctr.UUCP (Steven A. Dean) writes:
>BTW, does anyone know how fast data can be pushed through the parallel port

I've designed a hard disk controller that worked through the parallel
port and measured bursts of approx. 800k bytes per second - this was
using a 68010 with the transfer loop cached on the processor.  The
transfer is limited by the strobes on the PIA.

Mark Huth