Hi, 

> 
> Hi Cary,
> 
> <<WinPCap returns 47 bytes compared to EtherPeeks 64 bytes>>
> 
> I ran into this when I was first learning about WinPCap.  
> Turns out that for packets transmitted from the machine on 
> which you're running WinPCap, the padding bytes needed to 
> bring packets up to the minimum Ethernet frame of sixty bytes 
> (less the hardware-generated 4-byte CRC) are not logged.  So, 
> for example, if you run WinPCap and try doing a ping from the 
> machine on which you're running WinPCap, the captured ARP 
> request packets will show up in Ethereal as being 42 bytes 
> long.  Go to another machine on the same network and try a 
> ping, and the captured ARP request packets will show up as 
> being 60 bytes long.
> 
> Although I'm not intimately acquainted with the innards of 
> WinPCap, I've been told this effect is caused by the layer at 
> which the NDIS miniport driver intercepts the Ethernet packet 
> sending calls.

That's correct. From this point of view, winpcap just "relies" what NDIS
passes to its driver (which is a protocol driver). Turns out that, when
transmitting a packet, the padding is added at the underlying layers, and so
it's not received by WinPcap. The same for the CRC: NICs (or NIC drivers)
normally check and remove it before giving the packet to the protocols.

Loris 

 
> Rob---
> 
> 
> 
> 
> ==================================================================
>  This is the WinPcap users list. It is archived at  
> http://www.mail-archive.com/[email protected]/
> 
>  To unsubscribe use
>  mailto: [EMAIL PROTECTED]
> ==================================================================
> 




==================================================================
 This is the WinPcap users list. It is archived at
 http://www.mail-archive.com/[email protected]/

 To unsubscribe use 
 mailto: [EMAIL PROTECTED]
==================================================================

Reply via email to