Re: OT - sort-of
From: joel garry <joel-garry_at_home.com>
Date: Tue, 27 Mar 2012 09:26:58 -0700 (PDT)
Message-ID: <50e26a8c-dd25-493e-83c8-2ed33efd7869_at_z5g2000pbu.googlegroups.com>
On Mar 26, 5:02 pm, onedbguru <onedbg..._at_yahoo.com> wrote:
> On Sunday, March 25, 2012 1:46:04 AM UTC-4, Mladen Gogala wrote:
> > On Sat, 24 Mar 2012 19:38:16 -0700, onedbguru wrote:
>
> > > If you didn't get the packet, how do you know one was sent? (lol)
>
> > Because there is a packet counter and there is sequence number in every
> > packet. A hole in the sequence numbers means dropped packets. There is
> > something called "window". Peers enumerate packets in every window, with
> > the number of packets in the window typically being 256. There is a
> > special handling at the end of the window and if the receiver doesn't
> > acknowledge the entire window by the receipt packet, the entire window
> > will be retransmitted. This trick was introduced to avoid sending ACK for
> > every received packet, which was done a long, long time ago with RS 232C,
> > lovingly known as the "serial interface". Speeds were an issue back then,
> > so instead of acknowledging receipt of every received packet with a
> > received checksum, they started putting checksum in the packet itself
> > and, with the communication becoming more reliable, acknowledging large
> > group of packets. That was also implemented in an ancient protocol known
> > as X25, which was used among the Greeks during the siege of Troy.
> > When very fast and reliable communication was created, the packet and
> > windowing optimization was kept. The "very fast and reliable
> > communication" were the first Ethernet and the differential version of
> > the serial interface, RS-422.
> > Funny thing is that windows and packets also exist in IPv6.
>
> > --
> >http://mgogala.byethost5.com
>
> You must be thinking of RUDP - not UDP. UDP has no "sequence" number - that is why it is called "fire and forget" because there is no re-transmission of a packet. It is connection-less protocol that provides a simple and UNRELIABLE message service. If it does handle any network glitch, it would be handled at the application layer, not the network layer. The header contains no such "sequence number". Obviously the "joke" proved the protocol, because apparently you didn't get it :) :) :)
>
> There may be some ERRONEOUS information on the "web" that may suggest that UDP does do windowing. It does not. They may infer that UDP is a connection-oriented protocol. It is not. UDP could be viewed as "broadcast-only" protocol - no windowing and no acknowledgement (ACK).
Date: Tue, 27 Mar 2012 09:26:58 -0700 (PDT)
Message-ID: <50e26a8c-dd25-493e-83c8-2ed33efd7869_at_z5g2000pbu.googlegroups.com>
On Mar 26, 5:02 pm, onedbguru <onedbg..._at_yahoo.com> wrote:
> On Sunday, March 25, 2012 1:46:04 AM UTC-4, Mladen Gogala wrote:
> > On Sat, 24 Mar 2012 19:38:16 -0700, onedbguru wrote:
>
> > > If you didn't get the packet, how do you know one was sent? (lol)
>
> > Because there is a packet counter and there is sequence number in every
> > packet. A hole in the sequence numbers means dropped packets. There is
> > something called "window". Peers enumerate packets in every window, with
> > the number of packets in the window typically being 256. There is a
> > special handling at the end of the window and if the receiver doesn't
> > acknowledge the entire window by the receipt packet, the entire window
> > will be retransmitted. This trick was introduced to avoid sending ACK for
> > every received packet, which was done a long, long time ago with RS 232C,
> > lovingly known as the "serial interface". Speeds were an issue back then,
> > so instead of acknowledging receipt of every received packet with a
> > received checksum, they started putting checksum in the packet itself
> > and, with the communication becoming more reliable, acknowledging large
> > group of packets. That was also implemented in an ancient protocol known
> > as X25, which was used among the Greeks during the siege of Troy.
> > When very fast and reliable communication was created, the packet and
> > windowing optimization was kept. The "very fast and reliable
> > communication" were the first Ethernet and the differential version of
> > the serial interface, RS-422.
> > Funny thing is that windows and packets also exist in IPv6.
>
> > --
> >http://mgogala.byethost5.com
>
> You must be thinking of RUDP - not UDP. UDP has no "sequence" number - that is why it is called "fire and forget" because there is no re-transmission of a packet. It is connection-less protocol that provides a simple and UNRELIABLE message service. If it does handle any network glitch, it would be handled at the application layer, not the network layer. The header contains no such "sequence number". Obviously the "joke" proved the protocol, because apparently you didn't get it :) :) :)
>
> There may be some ERRONEOUS information on the "web" that may suggest that UDP does do windowing. It does not. They may infer that UDP is a connection-oriented protocol. It is not. UDP could be viewed as "broadcast-only" protocol - no windowing and no acknowledgement (ACK).
I've only heard third-hand that RAC uses UDP. The consequence is that whatever application uses UDP must handle things like retransmission, sequencing and error handling. Nothing wrong with that, if you think that having application programmers should be handling such things - personally, I think in most cases they shouldn't, but with the RAC development resources, perhaps there is enough expertise to Do It Right Eventually (DIRE).
But of course, I really don't know anything about it, and am now curious whether this rude p thing is involved, and whether it just gets too complicated for apps to handle at long distances.
jg
-- _at_home.com is bogus. I took the Disney Jungle Cruise yesterday, and at the end, the lady said: "And now here is a final joke for the mind readers and psychics."Received on Tue Mar 27 2012 - 11:26:58 CDT