[ih] Arpanet raw messages, voice, and TCP

[Vint Cerf]

On Nov 26, 2009, at 9:19 AM, Matthias Bärwolff wrote:

>
>
> Noel Chiappa wrote:
>>> From: =?ISO-8859-1?Q?Matthias_B=E4rwolff?= <mbaer at cs.tu-berlin.de>
>>
>>> I gather from the 1822 report .. that hosts could send uncontrolled
>>> messages (one packet messages at that) that would be delivered  
>>> without
>>> paranoid error control IMP-to-IMP and without RFNMs and  
>>> retransmissions.
>>
>> This is probably not very useful/interesting, but I looked at the  
>> MIT/Proteon
>> router ARPANet interface code (which dates from circa 1981/2, if  
>> memory
>> serves), and it had no capability to send Uncontrolled (type 0,  
>> subtype 3)
>> mesaages. (It would correctly process any it received, though.)
>>
>
> Thanks, this is by no means "not very useful/interesting", in fact  
> it is
> one of the only data points on the issue I have yet come across.
>
>>> What experiments or actual applications did people do with the raw
>>> messages?
>>
>> Note that Uncontrolled messages could not be longer than 1 IMP-IMP  
>> frame
>> (which was 1008 _bits_, IIRC). That's because to send a message  
>> longer than 1
>> IMP-IMP frame, the sending IMP had to first allocate a re-assembly  
>> buffer at
>> the destination IMP. Doing so would obviously be in conflict with  
>> the whole
>> goal of Uncontrolled messages (fast/cheap transmission), but a  
>> maximum packet
>> size of ~120 bytes would obviously not have been that much use for  
>> TCP/IP in
>> general.
>>
>
> I don't understand this. Even the 1974 Cerf/Kahn specification of TCP
> knew of "breaking up messages into segments", because "the local  
> network
> may limit the maximum transmission size" and which are, in turn,
> packaged into an internetwork packet. While hosts were eventually
> expected to accept IP packets of at least 576 bytes, they sure can  
> cope
> with smaller packets, too (even, say, just the header). Why then would
> 126 bytes foreclose experimenting with TCP/IP?
There was a minimum size IP packet - 576 bytes - and that was too big
for the 1008 bit IMP packet size. So we did NOT try to run uncontrolled
messages in support of TCP/IP.
>
> I am aware that the uncontrolled mesages could not be multi-packet
> messages. And, unlike ordinary single-packet messages they would not  
> be
> subject to the source-destination-IMP ack/retransmit facility. This is
> precisely the reason why I would sort of consider them an early  
> variant
> (or at least a conceptual anticipation) of the later IP packets. And,
> surely it must have been more fun playing around with higher level
> end-to-end reliability in a network that actually does drop a packet
> sometimes.
we had no problem with that: packet radio, packet satellite and  
ethernets
as well as Bill Plummer's Flakeway gateway provided us with plenty of
opportunity to deal with packet loss.
>
> It would really by interesting to learn more about the ICCB/IAB versus
> BBN clash on dropping TCP onto uncontrolled Arpanet messages, as  
> alluded
> to in the earlier Bob Braden email.
I don't remember this "clash" as well as Bob seems to, so can't  
comment further.
>
> Matthias
>
>> 	Noel
>>
>
> -- 
> Matthias Bärwolff
> www.bärwolff.de
>