[nsp] atm encapsulation overhead

[Phillip Vandry]

On Tue, Nov 18, 2003 at 09:41:20AM +0100, Iva Cabric wrote:
[...]
> 
> You can find information about overheads in different protocols
> (Ethernet, ATM, POS, ...) here:
> 
> <URL:http://sd.wareonearth.com/~phil/net/overhead/>

I may be wrong, but isn't the description of HDLC byte stuffing incorrect
in this document?

It says: The flag is 0x7e, and bytes with values of 0x7e or 0x7d in the
payload are escaped with 0x7d. This is the way it works with asynchronous
PPP (like on a modem or RS232 line). In this case POS is being described,
so I believe the bit oriented (not byte oriented) HDLC algorithm should
be the one given:

Six consecutive 1 bits mark the end of a frame. A zero bit after six or
more consecutive 1 bits marks the start of the next frame (Note that
0x7e meets this). In the payload, a zero is inserted after every group
of 5 consecutive 1 bits in order to be certain that 6 consecutive ones
never appear in the payload.

The pathological case for async PPP (data all 0x7e's) is not nearly so bad
with sync (HDLC) PPP because of this.

-Phil