Sorry for the long time to reply to your message, but I had to wield
my sthethoscope the whole weekend and was quite tired.
First of all I would like to congratulate you on having successfuly
conquered the Millenium bug, which I conclude from the fact that you
managed to reply three days (April 11) before my message was written
(April 8) :-)-O.
In message <[log in to unmask]>, "Anthony J. Brooks" writes:
> There is a contradiction is this response that I would like to point
> out and clarify.
I must confess that even after several days trying I am not able to
understand what contradiction you could mean.
> Here is where the contradiction begins.
> >If you dial up to an ISP, using PPP you are given a UNIQUE IP address
> >temporarily. Your computer is now visible on the Internet. The TCP
> >protocol establishes a virtual, error free connection with the target
> >software. It does not mind about where that software sits, IP does
> >that part.
> AH! If the computer didn't need an unique IP address, then, why is
> it imperative that the ISP assigns one to a dial-up computer - never
> mind whether the dial-up protocol is PPP or SLIP.
Trying to make sense of this gives me headache.
Either you overlooked the word "temporarily" or you are mixing up the
TCP and IP protocols.
> The fact is as you say, ANY device (computer, router, gateway, ...)
> that is communicating on the Internet (whether it is for 1 minute or
> several days) must have a unique IP address. And it doesn't matter
> if that IP address permanently resides with that piece of equipment.
If I read this correctly you agree with what I am saying about IP
> The thing that does matter is that the IP address unique identifies
> the communicating host in the Internet.
...during the time such device is connected to the Internet.
In other words as soon as you put doen the phone someone else might
get that particular IP number.
> DHCP (Dynamic Host Configuration Protocol) is one of the mechanisms
> that ISPs use to dynamically assign IP addresses to dial-up host.
I do not understand one little bit why you would bring up another
complication on the technical points of how to allocate IP numbers.
DHCP is of no relevance to this discussion. In fact, using DHCP for
this exposes you to the need of having a server up in order for the
dialups to get the IP numbers (can be very painful).
The net effect is in any case, that each dial-up device (modem or
rather port on the router or terminal server) has a unique IP number
either permanently or temporarily ascociated with itself, which then
temporarily is used by the computer doing the dialup (via PPP).
> Let's dissect the composition of the email address that Jeff ask about:
Again you are mixing the different protocols (layers) up.
An email address has *NOTHING* (yes I am screaming) to do with IP
numbers or domain names. And your "analysis" of the email address is
The Domain Name System works on the Application Layer (eg on top of
TCP, in fact it uses a different protocol called UDP instead of TCP)
to translates Domain Names into IP numbers.
However, that doesn't help us much, because you can not just send
emails to IP numbers, in particular not through a gateway (or
Therefor the DNS has several types of records, one, let's call it type
"A", translates a name into an IP number. Then there is the "CNAME"
which for example aliases the many WWW.something into proper Domain
Names for which there are "A" records. And there is the "MX" record
which points towards Domain Names for which "A" records exist. And of
course there is the "PTR" type which is used to reverse the process,
get a Domain Name from an IP number.
Now, an email address is interpreted by the Mail Transfer Agent
(sendmail, also on the Application Layer) through a call to what is
commonly called a "Resolver Library". This sends requests to DNS
servers receiveing the IP numbers of MX hosts for the Domain Name.
(Never mind whether the Domain Name is generic [log in to unmask] or
specific as in this example, MX records are returned.
The MTA then tries to talk via SMTP (over TCP) to each MX starting
with the one closest to the target until it succeeds to connect and
delivers tha mail to that MX (any farther MX will try to deliver to
The closest MX can be the system on which the mailbox sits or a
gateway or firewall.
> So this person "cbnrm" has a unique IP address (although it maybe a
> temporarily one) to communicate on Internet (actually the host
> he/she is using).
This is absolute nonsense. In this particular case it is as I
speculated in fact a FIDO network.
> Let me say though the IP address (which is used primarily for
> routing packets though the Net or any network running TCP/IP for
> that matter) used by cbnrm may be that of a server host.
This is incorrect. cbnrm does not use any IP number, the DNS system
resolves the Domain Name to the nearest working MX's IP number.
> But on the Net (us outsiders), this person has a IP address and an
> email account on a host machine that access can be gained to via the
> Simple Mail Transfer Protocol (SMTP).
There is no requirement whatsoever for the end user to have an email
account on a machine that has an IP number (implied by the statement
that it can be reached by SMTP).
Only MX hosts need to be reached by SMTP.
I think your contribution up to this point can be summarized best by
this quote from a private message I received:
"If he had just spent a few minutes doing a DNS lookup on
eo.wn.apc.org he might have kept his fingers off the keyboard
and avoided the embarrassment that is now his due."
> Remember, IP routing is done at layer 3 (network layer) of the OSI
> ref model
Now we come to the part where I had to be physically restrained.
The words "Anglais assez piquant" were privately used by a
correspondent who is reading this list (for occasions such as this)
> and email (via SMTP) is processed at layer 7 (application layer) of
> the OSI model. (We data comm engineers are disciplined to think and
> communicated in terms of protocol layers. These models are used for
> communicating effectively. You have to have an understanding of
> these model (OSI reference model as is relates to TCP/IP in this
> case) to really understand how this stuff works.)
I can not agree more with you. It is rather sad, however, that you
yourself don't seem to understand any of it.
Maybe I should refresh your memory (from some lecture notes):
The Physical Layer has the function of providing the
Ethernet, Modems come to mind.
The Data Link Layer consists of drivers (also Ethernet
controllers) creating frames out of the raw bits receiving
from the Physical Layer and moving them to neighboring hosts.
The Network Layer provides the software routines for routing
data to destination. It must decide if a packet received is
for the local host or needs further transfer.
The Transport Layer can reliably transmit data between hosts
and provides flow control. It makes sure the system only
receives as much data at a time as it can handle. It may
retransmit if necessary. ("Virtual Circuits").
The Session Layer manages reliable sessions between processes
by creating, coordinating and closing connections.
The Presentation Layer resolves differences in systems (like
The Application Layer is located at the top of the OSI
model. It deals directly with the user or applications such as
electronic mail, file transfer and the Web.
SMTP, POP, DNS
> As for the debate on using the WWW, I think that is not a matter of
> technology, but, rather, one of economics.
Agreed. Or as the Fisheries biologists would say "Catch per Unit
> Using the web (which is a component of the "Internet" service as
> commonly called and known in Africa), can be very expensive. The
> reason is because the PTTs charge a per minute charge on telephone
> local or long distance calls. Its not like here in the States,
> where we can make unlimited local calls and stay on the phone for
> hours for one flat rate.
Maybe not even only economics. But the net effect is, it is just worth
one's while to search for hours to find something.
> BTW for you Internet users in the States, did you know that the
> RBOCs are lobbying the FCC to start charging per minute usage
> charges for Internet access over the local loop?
I think this urban legend was first mentioned to this list a year or
> So the web may not be the answer economically for a lot of Internet
> users in Africa (maybe not here either if the RBOCs have their
> way). :) Email might have suffice. But I do agree, using email to
> access the web is a pain and awfully awkward because you have to
> know something about constructing the query using the query
You might find that the less resources one has available the more
resourceful one tends to become, in fact the opposite of what you are
writing about query languages may be true.
On a personal note, I hope USAID will hire you for many consultancies
so I have something to flame about in the years to come. :-)-O
PS: I also note that Nemo has yet to provide me with the URL of a
medical web site that is useful to a Health Care Worker.
Dr. Eberhard W. Lisse\ / Swakopmund State Hospital
<[log in to unmask]> * | Resident Medical Officer
Private Bag 5004 \ / +264 81 124 6733 (cell) 64 461005(h) 461004(f)
Swakopmund, Namibia ;____/ Domain Coordinator for NA-DOM (el108)
Vice-Chairman, Board of Trustees, Namibian Internet Development Foundation,
an Association not for Gain. NAMIDEF is the Namibian Internet Service Provider.