How can I setup a PHP socket connection using MAC address instead of IP address?

angelIII,

Thanks for the quick response -- but I need to do this programmatically.

My PHP program needs to scan the devices every few seconds and continuously update a visual representation of the device status.

It needs to be able to recognize if DHCP has changed the IP addresses for any of the devices, and change that device's socket connection to its new IP address.      

I understand that.
but first things first: can you check out if any of the 2 above suggestions can make it into (your) php code.
means: first try with 1 device manually with code doing just that.
after that, we will think about how to make this happen for the global picture.

What is the VB technique you have?

RQuadling:

I originally assumed that the IPPacketInformation.interface property would contain the MAC address, but have been unable to verify that. This may not be any easier in VB than PHP.

angelIII:

I'll have a look at the script you suggested tomorrow morning and let you know what I find.

When I tried the command on my XP Home box, I got "The following command could not be found: dhcp". According to the "help" info for "netsh", the only allowable sub-contexts that can follow the -c switch are:

   bridge diag firewall interface lan ras routing winsock  
   

If you load up wireshark (an IP packet sniffer/analyser), you will see that the mac address is always the closest point away from you. I'm on a wireless link, so the mac addresses I see are my laptop's and my router. That's it. No more details in the actual packet received.

Wireshark shows the following details (for a random packet) ...

Frame 1 - Can see that the packet is an eth:ip:tcp:ssl packet - essentially an https packet - I've got GoogleMail open over https.
Ethernet II - The data in this element is actually the first 14 bytes of the packet. It contains the mac addresses of the source and destination - essentially the hardware route. That's all that is needed as the router will look after which port to use for the IP address routing table it is maintaining.

After that it is IP data (20 bytes) and then TCP data (20) and then the SSL data (1430 bytes).

That's it. All done.

So, unless you've got ways of tracking the packets through each of the routers from the source to you ...

Basically the answer is no. The mac address you get is always going to be for the local hop only.

you can either delete the question, or accept your own comment as answer...

If you don't want to set static IPs on each device, but have a list of MAC addresses, you COULD use DHCP to assign an address based on MAC address. This saves the manual configuration per device - some device's have no easy interface, so doing it in DHCP is probably easier.

How would that work?

Would I somehow tell DHCP a MAC address and an IP address to assign to it?

If not, how would I know which IP address DHCP decided to assign to the MAC address?

The idea is that you can have fixed IP addresses without the need to manual program each device - some devices require dismantling to do this or connecting via a serial port - often a PITA to do.

By telling DHCP to reserve the IP address for a specific mac address, the IP address is managed by the DHCP server.

If the reserved addresses are in a fixed range, then you can easily iterate the addresses.

Not really a full solution, but just a few comments which hopefully make it easier to set up.


Open box.
Remove device.
Read mac address.
Enter mac address for reservation in DHCP in reserved IP range.
Put device into use.

Is it any different to fixing the address in the device?

Yes.

You would STILL need to remove the address from DHCP to stop DHCP from giving the address out to another pc or printer or device which can handle DHCP.

I'm just saying let DHCP do its thing.

That would work, as long as the DHCP server has the ability to assign a specific IP address to a specific MAC address. (My Linksys router is my DHCP server,  but it doesn't appear to have that feature).

But I might as well assign the IP address to the device in my application program.
It's a simple command, and I already have to manually maintain an IP-to-MAC table in the application anyway.

If I let DHCP assign them, I have to maintain the IP-to-MAC association in both places, risking a careless mismatch.  

I assumed you had something a little bigger than just a router doing the DHCP.

"If I let DHCP assign them, I have to maintain the IP-to-MAC association in both places, risking a careless mismatch."

Not true if you had a DHCP server which allowed you to reserve static IPs for specific MAC addresses.

I use this for collecting usage data from our networked scanners and printer.

All scanners are in 10.50.10.x and printers are 10.50.20.x ranges.

As a new printer or scanner is added to the network, the engineer takes the mac address off the back of the device and logs it in the DHCP server as the next reserved address. Job done.

The engineer lets the stats aggregator code know about the new device (essentially a web form) which scans the reserved range for new IPs. The only place anything is entered is in the DHCP server.

But, at some level, an IP-to-MAC table is maintained.

It MAY be worth using a different DHCP server (software based maybe) if this mechanism is suitable.

NOTE: this is a long post, but it's NOT meant to argue with your comments. Instead, I hope this more complete explanation of the problem might trigger new ideas, or a further explanation of something I might misunderstand.

@@@@@@@@@@@@

In your scenario, you add a new printer  in Office 203 with MAC = "MMM".
The engineer tells DHCP to reserve IP address of 192.168.1.101 for the printer with MAC = "MMM". HIS job is done.

Six months later, a programmer writes an application which must print to the printer in Office 203. She doesn't know its IP address, but that doesn't matter because she knows its network name. She writes to printer "Office203_Printer". HER job is done.

But my devices don't have network names. I need to write to them with PHP's "socket_write" function. If I'm correct, the ONLY way to do that is to connect to the socket with a command like this:
 
     "socket_connect(<socket>, <device IP address>, <device port>)"

Within the application program, I don't know the <device IP address> for the device in Office 203.

I can fix that by maintaining an "IP-to-DeviceLocation" table which my application program can read. When I want to write to "Office203_Device", I can look up its IP from the table.

If I already have to maintain that table, I can add the MAC address to it. My application can then look up the new device's IP addresses by its MAC address and download the IP address to the device.

I realize I'm doing work that DHCP could do, but I have to have the application table anyway to know the physical location of a specific IP address. If I maintain IP-to-DeviceLocation in my table AND maintain IP-to-MAC in DHCP, I have to maintain 2 tables which can get out-of-sync.

I BELIEVE that the simplest solution is to
   1. Reserve a range of IP addresses (ONCE) in DHCP
   2. Maintain an application table with IP, MAC, and DeviceLocation
   3. When a new device is added to the network, first add it to the application table, then use the MAC address from the table to set the device's IP address to the one in the table.
   4. When my application needs to write to a specific device, it uses the DeviceLocation to look up the associated IP address, then writes to that IP address.

@@@@@@@@@@@@

If I've misunderstood something in the previous suggestions, please set me straight. If not, I hope this wordy explanation will trigger some better ideas -- I really don't want to ask my client to maintain an application table, but so far I don't see a way around it.