On Tue, 17 Dec 2002, Major Variola (ret) wrote:
> Seems I have to explain why IP packet routing is not broadcasting some
> more. Those of you who understand that postcards have one trajectory
> from you to me can skip this.
>
> My first post was a first-order Choate fix. This post is a second-order
> fix. I refuse to respond to the next gripe, where JC brings up quantum
> postcards that take all paths at the same time, until you open your mailbox.
Yada yada yada...same old CACL bullshit.
> At 07:12 AM 12/17/02 -0600, Jim Choate wrote:
> >On Mon, 16 Dec 2002, Major Variola (ret) wrote:
> >> "The network?" Sorry, its one wire from here to there.
> >
> >No it isn't, try a traceroute to a regular site that isn't over your
> >internal network over several days, why does it change?
>
> In a *virtual* connection, the *physical* paths may change
> transparently.
Transparently says you, change the rules in the middle of the game and
hope nobody notices.
Thank you for making my point. One must have a physical connection prior
to a virtual connection. That physical network connection is equivalent
for this comparison to the physical connection between radio transmitter
and receiver, which is also shortest path (usually). That phsical
connection will change based on many variables. It is true that more
intelligent routers will cache various pieces of data, and provided the
cache doesn't go stale your route 'from here to there' will stay the same.
The reason that the intelligence was put into the routers was because the
packets were bopping around the network until their TTL went to zero
(each individual packet gets it's TTL decremented each time it hits a
router, until it hits zero when it's dropped, each router either sends it
to a known host on its local net or it's default route - where the process
starts all over again on that adjacent physical localnet).
The comparison to radio and multi-path distortion is also valid with
reference to receipt of multiple copies of a packet (and how prey tell
does that happen? Does the single router send out the same packet twice?
Nope, Different routers send them out and they get to the recipient who
takes them based on first come, first served -by different intermediate
paths-).
Bottom line, if there are n hosts on a network link and a packet is
injected each host gets a shot at it. If the host has sufficient info it
can make intelligent decisions, otherwise it drops back to the TTL so
the network doesn't get completely clogged by stale packets floating
around in limbo for perpetuity.
> Each IP packet has one path though the sequence of packets may take
> different routes.
Gibberish.
> Perhaps the mailing-postcards analogy is better than the telco one,
> since Ma Bell doesn't diddle the route after call setup AFAIK. But
> your postcards, once injected into the Postal Network, may take different
> routes. Not that you or your recipient knows.
No they won't. If you drop your postcard in a specific drop point then it
will be picked up and delivered to a specific central routing point. There
it will be collected with others of a similar destination. Then it will be
sent to the appropriate distribution center for that region. From there it
will be sent via truck or air to another distribution center, where the
reverse process takes place. About the only variance is the plane/truck
that is travelling the route between regional distribution centers
probably isn't the same one that took yesterdays mail, but it could be.
The USPS doesn't want your mail being sent all over hell and half of
Georgia, that costs us all way too much money.
> >Nobody (but perhaps you by inference) is claiming it is identical,
> >however, it -is- a broadcast (just consider how a packet gets routed,
> >consider the TTL for example or how a ping works). Each packet you send
> >out goes to many places -besides- the shortest route to the target host
>
> >(which is how the shortest route is found).
>
> Modulo CALEA and multi-/broadcast packets, each postcard is handed
> off to exactly one other device, or dropped.
Actually it's not. Take for example when my ISP send my packet (say this
email for example) out on their T3 or SONET link, there will be MANY
other hosts who will look at it and their inbound routers will try to
route it, unless they happen to know that destination IP is not in their
domain. Once the packet gets on a backbone -many- potential routes see it
and decide to pass it on to their default routes or drop it based on the
routing table and protocols (which are not spec'ed by TCP/IP). This sort
of broadcast is also why Ethernet itself uses the collision detection and
resend the way it does. It's also why Ethernet gets bogged to near
uselessness when the actual network bandwidth load approaches 50%.
This is analogous to tuning your radio to a specific frequency (ie IP
= frequency; protocol = modulation technique). The other issues that you
raise are -really- strawmen.
--
____________________________________________________________________
We are all interested in the future for that is where you and I
are going to spend the rest of our lives.
Criswell, "Plan 9 from Outer Space"
[EMAIL PROTECTED] [EMAIL PROTECTED]
www.ssz.com www.open-forge.org
--------------------------------------------------------------------
