| |
PA2AGA > TCPDIG 30.08.96 04:32l 205 Lines 7101 Bytes #-10903 (0) @ EU
BID : TCP_96_174C
Read: GUEST
Subj: TCP-Group Digest 96/174C
Path: DB0AAB<DB0PV<DB0WGS<OE3XSR<OK0PBX<OK0PKR<OK0PBB<OK0PAB<OH3RBR<OH2RBI<
OH7RBA<PI8VNW
Sent: 960829/2013Z @:PI8VNW.#ZH2.NLD.EU #:57828 [Hoek v Holland] FBB5.15c
From: PA2AGA@PI8VNW.#ZH2.NLD.EU
To : TCPDIG@EU
Received: from pa2aga by pi1hvh with SMTP
id AA16425 ; Thu, 29 Aug 96 19:49:54 UTC
Received: from pa2aga-1 by pa2aga (NET/Mac 2.3.61/7.1) with SMTP
id AA00003651 ; Thu, 29 Aug 96 20:55:33 MET
Received: from pa2aga-1 by pa2aga-1 (NET/Mac 2.3.61/7.5.4) with SMTP
id AA00007987 ; Thu, 29 Aug 96 20:40:07 MET
Date: Thu, 29 Aug 96 20:29:49 MET
Message-Id: <tcp_96_174C>
From: pa2aga
To: tcp_broadcast@pa2aga-1
Subject: TCP-Group Digest 96/174C
X-BBS-Msg-Type: B
BK> didn't turn it up in a quick grep of my bookshelf here at home.
It is probably possible to do this with PLD-type devices, but would be quite a
lot more work than straightforward microcontroller-based approaches.
BK> This is something you could throw a PIC uP at quite easily.
BK> Up, SRAM, and a lithium battery ought to do it quite nicely.
A PIC was actually the device I had in mind to do this.
-- Mike
------------------------------
Date: Tue, 27 Aug 96 12:15:56 EDT
From: "Tom Moulton" <tmoulton@ram.net>
Subject: Looking for an RFC
Was an RFC ever published to cover the IP fragmentation that is used by NOS?
(ie usage of PID 08)
***********************
Thomas A. Moulton, W2VY
w2vy@ram.net
------------------------------
Date: Tue, 27 Aug 1996 17:19:27 -0700 (PDT)
From: Phil Karn <karn@qualcomm.com>
Subject: Looking for an RFC
>Was an RFC ever published to cover the IP fragmentation that is used by NOS?
>(ie usage of PID 08)
No, that's an internal feature of AX.25. I think it's in the AX.25 spec.
As I recall, it came from Eric Scace.
Phil
------------------------------
Date: Tue, 27 Aug 1996 02:18:07 -0700
From: brian@nothing.UCSD.EDU (Brian Kantor)
Subject: PPP/SLIP in connected AX25?
Karn>A Motorola BitSurfer and an ISDN line.
BitSurfRs are $260 mailorder in the USA. PacBell wants $200 for the
ISDN installation and $24 for the first month, which will give you 112
Kb/s or better. That's $500 with enough left over to pay the taxes AND
buy myself a nice dinner. If I promise to keep the service for two
years, they'll even waive the installation fee.
And that's why I'm not active in ham radio much any more either.
Of course, I do this for a living, which is why I think my employer
should pay for the whole thing anyway. Who needs a busman's holiday?
- Brian
------------------------------
Date: Mon, 26 Aug 1996 22:44:08 -0700 (PDT)
From: Phil Karn <karn@unix.ka9q.ampr.org>
Subject: PPP/SLIP in connected AX25?
>It produces abominable results unless the TNC is very very smart and
>sends packets broken on the SLIP/PPP framing boundary only. If you do
Seems to me the easiest way to handle this is to use the idle-time
packetization feature, where the TNC keeps collecting packets from the
serial interface until there's a pause. Since the SLIP/PPP packets
come out of the computer as a continuous burst, this should do the trick.
At least I *think* AX25 TNCs provide this feature -- right? It's been
a long time since I've used one. Maybe I'm thinking of X.25/X.28/X.29
PADs...
Phil
------------------------------
Date: Mon, 26 Aug 1996 22:21:37 -0700 (PDT)
From: Phil Karn <karn@unix.ka9q.ampr.org>
Subject: PPP/SLIP in connected AX25?
>Thinking about the mtu/paclen relationship. I suppose you could get
>away with setting setting mtu == AX.25 Window * PacLen ok. Finding
>the optimum balance between IP overhead and Too Large Window could
>be a fun experiment.
Actually, the MTU could be anything you want, as the TNC will chop it
up into paclen-sized frames over the air. Of course, the round trip
delay could be quite long at 1200 baud, meaning lots of spurious
retransmissions until the TCP inside the Winsock stack adapts...
>Ironically it would appear there is actually some advantage to be
>gained from using this system. That being that is will be possible,
>even likely at times, for more than one datagram to be carried in the
>same AX.25 frame, helping minimise AX.25 overhead.
Yes. Brian Kantor made this point quite some years ago, and in looking
back I think it would have been the better thing to do, at least if you're
going to run connected mode.
On the other hand, TCP works hard to send a few max-sized packets
instead of lots of small packets, so the gain for a single active TCP
connection may not be quite as great as you think it is. The payoff
would be greater between two routers that carry many simultaneously
active TCP connections.
Phil
------------------------------
Date: Mon, 26 Aug 1996 22:37:18 -0700 (PDT)
From: Phil Karn <karn@unix.ka9q.ampr.org>
Subject: PPP/SLIP in connected AX25?
>I strongly disagree with this view. PPP adds more data, but it isn't
>"overhead" unless it does no useful work. With PPP, you get more automatic
Right. One of the things I'd change if I did KISS over again is to add
a CRC or checksum on the frame. I assumed that a simple piece of wire
between the computer couldn't possibly make an error, so why have a CRC?
Well, the wire is reliable enough, but the damn UARTs on the ends aren't.
They drop characters all the time (or at least they used to before the
days of 16550 chips on 133 MHz pentiums). PPP has a frame CRC, SLIP doesn't.
>I think this is a the real question. The more thought I give to Phil's
>proposal, the more I realize that it is completely at odds with what Phil has
>always proposed in the past.
Well, yes. It's just a hack intended to support those with a single
computer running Windows 95. Of course, the bigger question is whether
the data rate provided by existing TNCs can possibly be considered useful
with modern applications like Netscape...
Phil
------------------------------
Date: Mon, 26 Aug 1996 23:00:52 -0700 (PDT)
From: Phil Karn <karn@qualcomm.com>
Subject: PPP/SLIP in connected AX25?
>Heh. I profess to have been baffled by the proposal too. I'm pretty sure
>Phil wasn't intending the stuff to go via radio in the fashion described
>though. I interpreted Phils message to be solving the problem of how to
>get Win95/Win3.1 users to use their normal Win based winsock
>applications
>and interface them to NOS without needing to have new software written.
No, that would mean they'd need a second local machine to run NOS and
route IP over AX25. That is indeed the preferred configuration. But if
you have that configuration, use Ethernet for the inter-machine link;
no need to bother with SLIP or PPP over local hardwired lines in this
day and age.
Phil
------------------------------
Date: Tue, 27 Aug 1996 07:23:12 -0400 (EDT)
From: "Brian A. Lantz" <brian@lantz.com>
Subject: PPP/SLIP in connected AX25?
On Mon, 26 Aug 1996, Phil Karn wrote:
> There were (and are) several very good reasons for putting AX.25 into
> NOS:
I'd add one more:
7. It makes it easy to experiment with extensions to the AX.25 protocol.
Tom Moulton and I have been working on a way to add various features to
AX.25, and NOS has made a great platform for that. While I *DO* have the
tools to do Z80 eprom work, it is NOT a great way to do that kind of
To be continued in digest: tcp_96_174D
Read previous mail | Read next mail
| |
