Setting up AppFS on an Econet (HOWTO)

discuss general risc os software applications and utilities
Related forum: adventures

Post Reply
Posts: 626
Joined: Fri Apr 04, 2014 6:42 pm

Setting up AppFS on an Econet (HOWTO)

Post by philpem »

Angelsoft AppFS has been downloadable from Chris's Acorns for a while, but I don't think I've ever seen a scan of the manual. I decided to have a play with it this evening, and it turns out to be quite easy to set up.

First the big gotcha: AppFS works by transferring data on the disc sector level. It shares the contents of an entire hard drive (or partition if your FS supports them).
This means that If you point AppFS Server at a RISC OS 4 format (with long filenames) disc, you'll only be able to access that share on other RISC OS 4 machines. If you share an older-format disc, anything RO3+ will be able to access it. This is entirely theoretical of course, as I only have one RISC OS 4 machine (the SARPC).

Setting up the client and server
Unpack the zip file into a convenient place.
Shift-double-click !AppServer to open the application, then open !Run in a convenient text editor.
Search for this line:

Code: Select all

Appserver_SetUp -f adfs -d 4
The "-f" parameter is the filesystem name (e.g. IDEFS) and the "-d" parameter is the drive number. There are probably more parameters, but without a manual I don't know what they are...

Shift double-click !AppFS, then open !Run in a text editor.
Search for this line:

Code: Select all

AppFS_EconetSetUp -s 0.254
Change "0.254" to be the station number of the AppFS server.

Starting the server
Run !AppServer on the server. By default, this will also start !AppFS. (set "AppFS$UseServerAsStation" to "FALSE" in !AppServer.!Run if you don't want to do this).
You should be able to browse the AppFS by clicking on the AppFS icon.

Starting the client
Copy !AppFS onto the client machine (or Level4 share) and run it.
Click the AppFS icon on the icon bar and you should be able to browse the AppFS share across the Econet.

First impressions
It's certainly a lot faster than standard Level 4. PenDown started much faster. A configuration interface like Level 4's would have been nice, but it is what it is: a read-only FileCore filesystem sharer.
I've read that it supports AUN too, but I don't have a machine to test that with.

I'm curious how it stacks up against Acorn's Broadcast Loader and Application Accelerator -- sadly I don't have copies of these (they came with Level 4 Issue 3).

This should get you started -- now, does anyone have a scanned copy of the manual? :D

Posts: 85
Joined: Tue Oct 28, 2014 9:54 pm

Re: Setting up AppFS on an Econet (HOWTO)

Post by richw »

I used it back in the day, on Econet and Ethernet.

There certainly are more command line arguments. You can configure ram caching and perhaps some network transfer options. I wish I could remember where I found them! I seem to recall there were two or three major releases, maybe v2 was DCI4 compatibility?
Posts: 626
Joined: Fri Apr 04, 2014 6:42 pm

Re: Setting up AppFS on an Econet (HOWTO)

Post by philpem »

Been playing with this some more. It really doesn't like RISC PC Filecore, Simtec IDEFS, and especially RISC OS 4. I have a hunch it can't handle partitions over 500MB.

I've found a little more about the command line parameters: ... BMgimTdSAJ

Code: Select all

In your earlier message you wrote:

>Is there anyone out there who uses Angelsoft AppFS 3.00 on an Econet
>(or even Ethernet, I suppose) network and would let me ask them
>for some info from the manual?

>I need some information on the cache settings. They have 4Mb servers

>which have over 2Mb of RAM doing nothing and looking at the AppFS modules
>it seems possible to configure cache options. How?

In the !RUN file - in !Appserver you will find the appserver_setup lines

|activate the server followed by 'Econet' comms module
Appserver_SetUp -f ADFS -d 4 -n 800 -b 8192
AppServer_EconetSetUp -b 34

This means that ADFS drive 4 is being exported with 800 packets of 8192 bytes
of cache. (Use 4096 bytes for Econet! - especially if you have Acorn bridges -
I think SJ bridges will cope with 8K packets) I therefore have about 6M of cache
and its on a RiscPC - as pre RPC stations cannot have more than 4MB of cache
as all the cache has to be in RMA - so you probably couldn't have more than
about 2MB of cache anyhow considering the amount of module space used up

The Econetsetup line with -b 34 means that I have 34 open receive blocks which
cannot therefore cope with requests from 34 clients at a time. Typically, I am
told that requests are dealt with so fast that this only needs to be 3 or 4; but
I found with older server stations cannot keep up - perhaps I could reduce it
now on the RPC.

You can put a -s parameter in this line too - however, this allows recursive
calls to be made to the server which really makes it fly and is much better at
allowing the station to be used as a client station or for some other task


It is also prone to making the server lockup, which is a great shame.

Really APPFS needs an overhaul to cope with New filecore, dynamic areas rather
than RMA, and it needs to be fixed so that it can be re-exported at gateway stations
with cache so that pressure on the backbone is relieved.

Oh- it doesn't work with DCI4 very well either.

Alan Davies
Head of Computing
Birkenhead School
And another from ... Nc-JxToWAJ:

Code: Select all

In an earlier article David Williams <> wrote:
>In article <>,
> wrote:
>> A 25MHz A5000 tends to stall.... Access is the fastest user
>> protocol that I have used......APPFS is I think as good now provided
>> you enable the 'background' Start Econet TX commands rather than the
>> 'Do calls' (its not documented)...
>Can you expand on the AppFS setup Alan?
APPserver is started from its RUN file (which also contains all the
configuration separate configuration utility on this

There are two critical lines:

Appserver_SetUp -f ADFS -d 4 -n 200 -b 8192
AppServer_EconetSetUp -b 34 -s

The first is well documented - exporting a filing system ADFS in the above case - drive4
with 200 blocks each 8K in length acting as a cache withing RMA space (limited to 4M
in total even on 8MB RO3.1 machines.....but up to 10MB I thing on RO3.7.......of course
about 2MB goes with modules softloading.....

I would try to ensure that there is enough cache space for client machines to boot...start
up one client.....if the disc light flashes on the server when you start the second increase
the size of the cache. On most setups this is large enough to ensure that most applications
will cache as well.

The next line opens 34 receive blocks. In theory 2 blocks might be enough in that
when one is being dealt with the second block will fill with the next request....and
so on was it not for....

the -s command (undocumented?) which enables the application server to multitask can call
its transmit commands and proceeds to deal with other requests without waiting while the
transmit finishes. This instantly 'free's up the desktop on the server machine very
noticeably on later models but also brings benefits with 25MHz Arm3 A5000 stations.

ALthough this has always been possible with ANT Ethernet cards (In my experience) it has
only been reliable with I-cubed ethernet cards since EtherH4.16 ( my experience)

(On the negative side we have a whole load of ANT cards - Ether3 and EtherB which
are really awful using Access via any of the repeaters we have on site...although
EtherM and Icubed cards have no problems.....)

We have 30 machines in one room and using 34 blocks guarantees that no machine is left out -
with retries attempting to get its request to the server (stil leaving some capacity
for other machines). With the -s flag you reduce to around 6 open blocks with little
attempt on load times - although in theory there might be some increase in collisions and
network load.

We only use our Main APPFS server for a minimal boot now - transfering control as quickly
as possible to Omniclient (still not available in ROM despite all the hype over recent years)
and therefore onto our NTserver (running with Fast Ethernet at 100MHz feeding many separate
10MHz links via a switching hub). I remain EXTREMELY impressed that the NT server
can seemingly take everything we throw at it (Dual PPro 200MHz, 128M RAM) without
so much as flinching.......apart from some ocassional irritating timeout errors on client
stations...which don't appear entirely load related.

Alan Davies
Head of Computing
Birkenhead School
Interesting curiosity I guess, and fine if your fileserver runs RO3.1x it's fine... but it's pretty unstable. I'm not sure if it's down to trying to run AppFS on a RO4 RISC PC, but it seems especially good at crashing the machine to the point where I have to hit Reset or cycle the power.

Post Reply

Return to “32-bit acorn software: other”