"The Linux Gazette...making Linux just a little more fun!"

Programmer Fights with Subnets

From Grant Murphy on Tue, 17 Nov 1998

I'm a numerical C programmer and have inherited the system admin job in a 'small' geophysical exploration company. We have a fine collection of lovingly maintained and oftem overhauled equipment ranging from SunOs4 machines to an NT box, handbuilt aquisition systems mounted in aircraft, dual real time differential GPS systems etc. etc. I know A LOT about a number of particular things in maths, geophysics, unix, world coord systems etc, but I am a babe in the woods about other things ... networking in particular.

The problem at hand ( & one that I have searched for FAQ's on & trolled comp.os.linux.networking for the REAL answer to ) is this:

We have two networks in our office, one is made up entirely of windows 95 machines and office printers etc. The other was made up entirely of SunOS4 and Solaris machines with an A0 HP map plotter and a versatec plotter ( about the size and weight of a compacted VW bug ). The two networks intersect in a single linux box running a 1997 version of caldera linux, with two network cards, a dial out modem card for internet access, no keyboard, no moniter ( well, who needs them )

The SunOS network now contains two windows machines used for processing data. One is Win95, the other WinNT workstation.

I **can't** get the two windows machines to see the shared drives and printers of the win95 machines on the other side of the linux box.

1) I have all win machines using TCP/IP with NetBeui disabled (lots of people seemed to recomend this)

That's because NetBIOS/NetBEUI (the "native" Windows transport level networking protocols) aren't routable --- they only work within a LAN).

2) I have samba on the linux box and can mount unix drives and see them on the network neighbourhood of the win95 box & winNT box on the unix network.

What version of Samba is it? Have all the appropriate patches and service packs been applied to the Win '95 and NT boxes?

That problem probably related to the share "browse mastering" protocols used by SMB. There have been many problems with these browsing protocols. I don't know the details, but I've heard that the Samba team has done quite a bit of work to fix those problems.

3) The network was split into two rings before I arrived under the rationale that the traffic of the two networks wouldn't interfere (some of the geophysical data traffic is pretty big - half gigabyte files etc)

Isolating LAN segments is a classic and effective way to optimize bandwidth utilization. I shudder t think of the amount of money that's been unecessarily and poorly spent on etherswitches for networks that would have benefitted far more from simple segmentation and (in some cases) some server file tree replication.

4) The linux box has two cards: eth0 with IP address 192.9.200.10 and broadcasting 192.9.200.* - all unix boxes, win machines attached through that card have IP addresses 192.9.200.* eth1 with IP address 192.168.1.10 and broadcasting 192.168.1.* - all office machines have adresses 192.168.1.*

5) I can ftp from the office network to the unix boxes alright .

So, TCP routing works between the two.

I'm under a reasonable amount of pressure to make the network look easy, people want to access the HP A0 plotter from the office computers just like they access the office laser printer - Now that the processing guys have an NT box with word processing etc. they want to access the office laser printer.

If the primary resource that is to be shared is the printer --- I'd connect the printer to the Linux box, and install Samba. Let it be the print server as well as the router between the two segments.

Likewise for the plotter (if that can be driven by your Linux system. I'm not familiar with the device or its interfaces).

Owing to industry recession, the chances of getting an expert network guy in to solve it seem to be slim to bugger all. This is chewing up time that is better spent working on algorithms to do noise reduction of 256 dimensioned radiometric data, and improving field QC software.

If you have any answers to this conundrum they would be gratefully received & I am happy to return the favour with answer's to any posers that you might have about numeric/scientific/geophysical/C language problems.

Try installing the latest version of Samba on the Linux box (try the 2.0. beta that was announced last week). Hopefully it will be able to propagate those pesky browse/share broadcasts from each segment to the other.

(I wrote an ANSI C compiler for an early version of MINUX that was ported to both a transputer array and an ARM6 chipset machine - none of that involved networking though)

Is there any Linux support for transputers? Are there modern transputers (PCI, even), or have modern processors obviated their utility?

Yours sincerely (& perplexed)
Grant Murphy

Copyright © 1998, James T. Dennis
Published in The Linux Gazette Issue 35 December 1998

			office	largedisk	links	yamaha	magickeys
			passwd	ftproot	pvtmail	netware	crypto
			relay	project	bootmethod	sysadmin	ipscript
loopfs	mrtg	slimscan	rpm	modutil	libc	dell	remoteroot