"Paul Boddie" <[EMAIL PROTECTED]> wrote:
> Fredrik Lundh wrote: > > Paul Boddie wrote: > > > > > This is one of the more reliable methods since upon receiving a packet > > > "delimiter" the receiver knows that the data is complete. > > > > and for people who want <strong>RELIABLE</strong> and not just "at least > > not entirely unreliable", there's always: > > > > http://cr.yp.to/proto/netstrings.txt > > That's why I hedged my bets and put "one of the more" rather than "the > most" in that sentence. ;-) I've been using netstrings myself, although > I didn't know that they had that particular name. > > > (if you control both ends, there's hardly ever any reason not to use > > netstrings. they're trivial to generate from Python, and pretty simple > > to parse.). > > Indeed. I was not aware that putting the length in front of a data packet had been blessed with the name "netstring". It is almost obligatory to do something like this if you are using raw sockets as otherwise you never know where you are. What I tend to do is more complex and probably also has a name that I am unaware of - I "frame" packets between tildes - 0x7e - the SDLC/HDLC flag character - this requires me to escape tildes, and my "escape" char, to prevent them occurring in the message - I use the ordinary forward slash as the escape char, and alter the poison chars by exclusive oring them with 0xff to invert - after adding at least a bcc formed by the xor of all the chars in the package, (before the escaping), and where its important, I use message numbers too... - but then - I am paranoid... And yes I once wrote some x25 type stuff in z80 assembler... :-) - Hendrik -- http://mail.python.org/mailman/listinfo/python-list
