An idiot's guide to OmniClient

OmniClient is one of the things I’ve always struggled to get my head around, specifically configuring LanMan connections. Since the LanMan implementation was “recently” updated to support newer authentication methods, and we’ve got a new RISC OS 5 User Guide on its way, it seems like now is a good time to try and come up with an OmniClient guide which covers typical setups and any likely pitfalls.

I know that I briefly had the new version of LanMan talking to both my Windows 7 PC and my Linux PC, but I’ve upgraded the OS versions for both machines since then (now Windows 10 and a 2017 flavour of Ubuntu), and like the idiot I am I didn’t make a note of what the working connection details were. So now that I’m trying things out again I’m not entirely sure whether any failures are my fault, LanMan limitations, or the result of the servers using an entirely new set of authentication/protocol versions.

One thing that I don’t think I’ve ever seen happen is for available shares to automatically appear in OmniClient’s window – I’ve always had to enter the connection details manually, which is where most of the room for mistakes comes in. Possibly this is just a configuration issue (on either the server or the client), so some guidance on that would also be welcome.

I suspect that the easiest way of getting the information across would be via a series of screenshots, showing how to correctly configure a Windows/Linux/Mac/whatever server and how to configure the RISC OS client (omniclient configuration, and network configuration, if relevant – e.g. does the “local domain” matter?). Plus a list of any issues/limitations with each Omni protocol (e.g. LanMan only supports mounting a share, not a directory within the share)

When I get home I’ll have to have another go at connecting to both my Windows & Linux machines and see if I can find the magic combinations that make things work.

The biggest problem I have is remembering that ‘Directory path’ in the “Mount ‘Lan Manager’ Server” window is really ‘Share name’ Otherwise logging on should be straightforward. It is all case insensitive except the password – which if you are using a windows machine is the password you use to log on to your windows computer. I don’t use the “Logon ‘LAN Manager’ network” window. And it’s quicker to use the mounts→protocol→Lan manager option on the iconbar menu to avoid having to wait for the hourglass to stop when you click on the iconbar icon.

A couple of notes that might help.

Windows 10 updates can reset sharing settings, this site helped me a lot, http://www.bapfish.org.uk/networking.htm

Ubuntu 17.10 appears to have acquired a Sharing section in Settings which contains a (new?) File Sharing option. This caught me out!

One thing that I don’t think I’ve ever seen happen is for available shares to automatically appear in OmniClient’s window – I’ve always had to enter the connection details manually, which is where most of the room for mistakes comes in. Possibly this is just a configuration issue (on either the server or the client), so some guidance on that would also be welcome.

That used to be a function of the NetBIOS elements of the PC end supplying a response to a broadcast query. Later that shifted into NetBIOS over TCP so that things were routable and revolved around use of WINS for resolution (which is the probable origin of the wait for the timeout situation). Later changes moved the resolution into DNS IIRC.

The biggest problem I have is remembering that ‘Directory path’ in the “Mount ‘Lan Manager’ Server” window is really ‘Share name’ Otherwise logging on should be straightforward.

That I think is best addressed with a change to the build such that it actually displays as the required item rather than trying to give some clue about what is happening on the server end (which the users generally don’t care about or understand anyway). That said I’d say the biggest problem is people trying to share directory paths with a space in the name…

It is all case insensitive except the password – which if you are using a windows machine is the password you use to log on to your windows computer.

More accurately the username and password combination need to be matched to a username and password for an account that can be validated on the server (WinPC/Linux box) and specifically an account which has permissions to access that resource on the server.

This is very timely; just two days ago I was trying to connect to my Mac without success (SMB is enabled but I kept getting “Cannot make TCP/IP connection”). I can connect to my Windows server so I know that OmniClient itself works, and I can connect to my Mac from my Windows machine so I know that sharing is enabled. Some instructions on connecting to a Mac would be appreciated :)

(Edited with correct error message)

When I get home I’ll have to have another go at connecting to both my Windows & Linux machines and see if I can find the magic combinations that make things work.

Of course, everything now works perfectly first time.

(well, apart from the shares being found automatically – it looks like I’ll need to do a bit more tweaking if I want that to work)

http://www.phlamethrower.co.uk/misc2/omni/lubuntu.png – Lubuntu 17.04
http://www.phlamethrower.co.uk/misc2/omni/windows.png – Windows 10
http://www.phlamethrower.co.uk/misc2/omni/windows2.png – Things to check/change on Windows

Also very important is the “Save mounts” option in OmniClient ;-)

(well, apart from the shares being found automatically – it looks like I’ll need to do a bit more tweaking if I want that to work)

This relies on there being a master browser on the network. Without one of those then there are no broadcasts saying who’s who. There can be several, and Windows will typically nominate itself as one, which forces a browser election to find who’s the best server (for some value of “best”).

If you sniff the network traffic with several Windows servers you’ll see an endless tussle of browser elections. From LanManFS’ point of view that’s useful because there’s more broadcast traffic with machine lists in.

Being broadcasts you’ll also want to check everyone’s using the right (or at least the same) netmask. That’s the most common cause of ShareFS not showing available discs via Freeway too.

With Windows as a client you see the exact same thing going to “My Network Places” on a foreign network. You have to wait 30s+ until some of the machine names start to populate, or mash F5 a few times. This is because either a browser election was in progress or the name list hasn’t been broadcast recently.

The biggest problem I have is remembering that ‘Directory path’ in the “Mount ‘Lan Manager’ Server” window is really ‘Share name’

If you think of a UNC path \\machine\blah\file.txt it sort of makes sense because ‘blah’ in that context is a directory; had I used a URI instead //www.machine.com/blah/file.txt it’d have been more obvious. Plus it’s worth remembering OmniClient is not LanManFS, it’s for other protocols too where calling that field a share would be equally backwards.

which forces a browser election to find who’s the best server (for some value of “best”).

Microsoft says:

The following criteria determine whether or not a browser has won an election:

• If the election version of the browser is greater than the election version of the sender, the browser wins. If not, the browser uses the next election criterion. The election version is a constant value that identifies the version of the browser-election protocol. The election version is the revision of the browser protocol and is not related to the operating-system version.

• If the election criteria of the browser is greater than the election criteria of the sender, the browser wins. If not, the browser uses the next election criterion.

• If the browser has been running longer than the sender, the browser wins. If not, the browser uses the next election criterion.

• If none of the criteria above have determined the election, then the server with the lexically (alphabetically, including numbers and symbols) lowest name wins. For example, a server named “A” will become master browser over a server named “X.”

Of course they don’t mention that W95 was a bit bugged and tended to believe it had won any elections it participated in. ;-)

Of course they don’t mention that W95 was a bit bugged and tended to believe it had won any elections it participated in. ;-)

Well, to be fair, 95 is completely and utterly unsupported at this point!

Acorn did a similar thing; the RISC OS 3 PRM states near the beginning:

There are very few remaining users of Arthur, and we consider it to be obsolete. You should not worry about making your programs compatible with Arthur.

In view of this, we do not distinguish features and facilities that are available under RISC OS but not under Arthur. However, you will find most of the facilities of Arthur described in this manual, because they have been subsumed into RISC OS. If you need full details of how Arthur did things, so you can maintain old programs, you’ll have to refer to the Programmers Reference Manual that was released with Arthur. Don’t throw your old manuals away – keep them!

If the election version of the browser is greater than the election version of the sender, the browser wins. If not, the browser uses the next election criterion. The election version is a constant value that identifies the version of the browser-election protocol. The election version is the revision of the browser protocol and is not related to the operating-system version.

Except in as much as the browser service that generates this is tied to the OS/service pack/patch revision. So basically, closely related to the OS revision.

If the election criteria of the browser is greater than the election criteria of the sender, the browser wins. If not, the browser uses the next election criterion.

Roughly: MAC/IP IIRC

If the browser has been running longer than the sender, the browser wins. If not, the browser uses the next election criterion.

Actually restarting the browser service triggers an election but at the end of the argument it tends to end up with the original master

If none of the criteria above have determined the election, then the server with the lexically (alphabetically, including numbers and symbols) lowest name wins. For example, a server named “A” will become master browser over a server named “X.”

Have to say that in a domain with several hundred servers I’ve never seen that.
Mind you having the servers spread across 8 different subnets 6 DNS/domain servers and 2 WINS servers it isn’t something I’ve looked at too closely.

Of course they don’t mention that W95 was a bit bugged and tended to believe it had won any elections it participated in. ;-)

ROTFLAWM. “A bit bugged” – do you do stand up?

ROTFLAWM. “A bit bugged” – do you do stand up?

I recall when the best advice computer shops had to offer was “reformat your harddisc and reinstall everything”. At the time I couldn’t understand how any operating system could ever hope to be popular if that’s how one recovered from the OS messing itself up. I used W95. I updated to W95 OSR1½. Both were awful in ways I can’t write here. W95 would habitually blue screen if I dared to eject the CD-ROM by pressing the button on the drive, or coughed while it was printing, or…
It wasn’t until W98SE that we had a system that did what was asked of it. I found W98SE to be reliable, ME was a mess, XP I’m still using, Vista was quietly taken out back and shot in the head. That “every other one” pattern implies 7 should have been fairly good, 8 not so… But wait! Doesn’t that mean 10 is supposed to be one of the good ones? Did I miss a release? Or is W11 going to be so bad we’ll be begging the spybot to return?

That said, I think there is a reason why I use XP (Dark Ed) and like RISC OS. I think it’s because I consider desktop machines and mobile devices to be different and with different UI requirements. It seems the operating system creators are putting too much effort into merging them into one coherent “whole”; but the thing is I create when I’m using RISC OS or my PC (even if “create” is ripping a DVD!). I don’t do that with my phone. Writing this is about as creative as it gets. Otherwise I’m watching kittens in YouTube, Chihiro Onitsuka, various videos of people blowing up stuff (sorry, that bloke microwaving a pack of TNT doesn’t get old), streaming radio, etc. Or just reading ElReg and enjoying the comments. The two – mobile device and desktop, are used in different ways. I don’t want them to be the same.

Actually restarting the browser service triggers an election but at the end of the argument it tends to end up with the original master

That’s a very American outlook isn’t it? Pick a fight you are unlikely to win in the hopes that maybe you’ll get lucky (punk). That tactic worked so well in Vietnam, and Korea, Afghanistan, Iraq…..

Seriously, isn’t it a rather ass-backwards protocol? Wouldn’t it have made more sense to fire off some sort of “identify” packet to get a response from the available browsers (and know where to look for information) and only start an election if the newly started one had something to offer?

I mean – what the hell happens when you bring a network back up after a power failure? Does nothing useful happen for the first ten minutes due to the machines arguing with themselves over who gets to hold the pink dinosaur?

That “every other one” pattern implies 7 should have been fairly good, 8 not so… But wait! Doesn’t that mean 10 is supposed to be one of the good ones? Did I miss a release? Or is W11 going to be so bad we’ll be begging the spybot to return?

W8.1 (8.0 was such a pigs ear that 8.1 came fairly rapidly.

I mean – what the hell happens when you bring a network back up after a power failure?

I check the core routers and the access switches and if I’m happy with that stage I go and make a tea or coffee and have a peek at the records of high traffic on access ports when I come back. (Every port is monitored)
Then I check the connection speed to the internet by logging on to something essential but unlikely to be cached in the proxy (https://www.riscosopen.org/forum/) fits the bill :)

I was trying to connect to my Mac without success (SMB is enabled but I kept getting “Cannot make TCP/IP connection”).

Some Googling indicated that Apple changed the SMB server a couple of years ago and it’s apparently been broken since then. I’ve now shared a directory over NFS but I’m having no luck with connecting to that either!

Omni doesn’t include the NFS protocol by default. I grabbed OmniNFS out of CVS but it appears to depend on another module simply called “NFS” which I can’t find. Is it buried somewhere in CVS or is there some other way to get it? Even without it I’m getting a bit more success than I had with SMB as I can now see my NFS server in Omni, but can’t actually connect to it (SWI 410C2 not known, which is NFS_SetUser, presumably provided by the NFS module).

A quick search yielded a couple of recommendations for Sunfish. Alas, its best feature seems to be that it’s very good at crashing (at least under my 23-Oct-17 ROM).

Any ideas?

I recently downloaded !Sunfish from !Packman, and it kept crashing on my RPi3 (low vector 26Jul2017).

I updated it with the !Sunfish front-end from Chris Gransden’s site RISCOSPORTS

This fixed !Sunfish so it worked OK with my Synology NAS.

Ralph

Thanks; I’ll try that later.

But since this is an OmniClient thread, I suppose my question is this: Does NFS actually work in the current version of OmniClient?

And what would it take to make Sunfish into an OmniClient “plugin”?

The last version of Acorn NFS I saw in the wild was for RISC OS 3.1 IIRC…later, everyone used ImageNFS by WSS. Did ImageNFS integrate into OmniClient? I don’t remember.

Thanks; I’ll try that later.

It’s better than it was, but is still crashing frequently. But this isn’t a Sunfish thread so I’ll leave it for now!

I must admit to having similar frustrations with my Mac and getting it to speak NFS/Samba. Gave up in the end!

Back in the day, I used NFS a fair bit with RISC OS 3.6 to 4.0 talking to Linux and Windows (!) machines, so I know it’s possible.

Does anyone have any insight in LanManFS charset/encoding handling? I am currently investigating the state of play using German umlauts for filenames with CIFS/SMB (LanManFS) and NFS (Sunfish) vs. my Netgear NAS.

Samba docs seem to describe various options manipulating encoding in various stages. Is LanManFS an “old type” client (i.e. DOS/Win9x mode) or a “new type” client? Any experience with LanManFS vs. LanMan98 concerning filename handling?

I saw problems where LanManFS (and Windows!) did not show files created with Sunfish. I saw information on the web that only NFSv4.1 started to properly handle filename encoding, but various Linux clients seem to “just use UTF-8” with NFSv3 and everything works.

If anyone has already done investigations in that specific area, I would be delighted to hear about it!

Is LanManFS an “old type” client (i.e. DOS/Win9x mode) or a “new type” client?

Old type, I believe. Adding support for CAP_UNICODE is probably the best way of dealing with things (hopefully any server from the past 10 years will understand unicode!)

https://www.riscosopen.org/tracker/tickets/435 (see Sprow’s comment at the bottom)

I did a disc build a couple of hours ago and “installed” it. Omni is v2.27 (29 Jan 18). Am I totally misunderstanding what it is, or is it a multi protocol network filesystem frontend?

Just like always, all I see in “Network Servers” is “LAN Manager”. Same goes for Mounts > Protocols. Is there something I’m missing? Are there modules that need to be enabled somehow?

You need to enable modules by editing the Startup file; details are in the User Guide.

is it a multi protocol network filesystem frontend?

It is.

Just like always, all I see in “Network Servers” is “LAN Manager”. Same goes for Mounts > Protocols. Is there something I’m missing? Are there modules that need to be enabled somehow?

For some reason you need to enable any additional protocols manually by editing !Omni.Files.Startup.

Just a very quick note to say that I have just managed to connect my Iyonix to my Synology NAS using NFS. I am not (yet) sure which changes I made are necessary, and which are not.

However, my interest in NFS was in the hope it may offer a performance increase over using SMB with LM/LM98 … but so far it looks much worse … and it does not seem to handle filename,filetype so filetypes are lost.

It all needs further experimentation, and I will report back in due course.

filetypes are lost

Hmm… I just copied a directory of differently-typed files to my Mac’s NFS share, dismounted, reconnected, reopened the share and the filetypes were all preserved. I wonder why it’s not working with your NAS; perhaps MacOS is doing something clever with the old HFS type/creator codes.

Edit: It’s not using the HFS attributes at all; it’s appending ,xxx like you’d expect. That’ll teach me to actually look at what it’s doing before posting :)