Date   : Sat, 31 Oct 2009 15:30:31 -0000
From   : famrowland@... (Andrew Rowland)
Subject: Econet packet size

Obviously my bad memory or misunderstanding at the time :)

Andrew

-----Original Message-----
From: Phil Blundell [mailto:philb@...] 
Sent: Sat, 31 October 2009 13:34
To: Andrew Rowland
Cc: bbc-micro@...
Subject: Re: [BBC-Micro] Econet packet size

The packet size has always been variable.  OSBGET's poor performance on the
original NFS was nothing to do with packet sizes, it was caused purely by
the absence of any buffering at all in the NFS ROM (and hence the
requirement for a round-trip to the fileserver for every byte).
Even a modest buffer of 32 bytes or so would have made quite a difference to
its performance.

I can't immediately think where you might be getting the figure of 256 bytes
from; that doesn't seem to match up with any obvious protocol limitation.

p.

On Sat, 2009-10-31 at 11:09 +0000, Andrew Rowland wrote:
> When did Econet start using variable packet sizes? Back when I as 
> using Beebs on Econet it was fixed at 256B, IIRC. Hence the poor 
> performance of OSBPUT and OSBGET until that school in Cambridge -- 
> Newhall School? -- came up a proper buffered version, which Acorn 
> incorporated into the next release of NFS. I remember that was a 
> much-awaited upgrade! Barnsley LEA got the chips in bulk to upgrade all
the Econet stations in the LEA.
> 
> Andrew
> 
> -----Original Message-----
> From: bbc-micro-bounces+famrowland=freeuk.com@...
> [mailto:bbc-micro-bounces+famrowland=freeuk.com@...] On 
> Behalf Of Rick Murray
> Sent: Sat, 31 October 2009 00:25
> To: BBC MailList
> Subject: [BBC-Micro] Econet packet size
> 
> I remember recently a discussion on Econet packet sizes. Was idly 
> reading stuff on the MDFS server, and I came across this paragraph:
> --8<--------
> The econet is a low cost, moderate speed (50-300kBit/S on-net) network 
> for connecting together microcomputers. It allows one machine to 
> transmit a packet of data to another which has a suitable reception 
> enabled to receive that data. A packet contains at least one byte, and 
> at most 1280 bytes, of data.
> --8<--------
> 
> Source: http://mdfs.net/Docs/Comp/Econet/RISCiX.txt
> 
> 
> It is an undocumented Acorn document. RISCiX is an ARM based system 
> that was, I believe, a Unixy contender to RISC OS that never quite caught
on.
> The Rxxx workstations (looking a lot like souped-up A440s to me!). 
> That would date it late-80s. I would imagine the 1280 is likely a 
> limitation of the 6502-based machines, >1K is a lot in a 32K computer! 
> At any rate, this seems the Acorn-quoted value so reasonable to figure it
is reliable enough.
> 
> FWIW, skipping the ARM/RISCiX stuff, this document is a lengthy but 
> pretty complete introduction to the ins and outs of how Econet works.
> 
> 
> Best wishes,
> 
> Rick.
>