[news.admin] The rebirth of USENET

mark@cbnews.ATT.COM (Mark Horton) (06/14/88)

>  Yesterday AT&T announced that ihnp4, cbosgd and att would be severing all
>  outside links and discontinuing third party mail.

Please understand this is not what we are doing.  We are not cutting
our outside links.  In fact, it was a strong need to keep our outside links
running that led us to decide to stop passing third party email as the
best way to cut costs.  Mail into and out of AT&T through the "att"
gateway will continue.  "att" is serving as a professionally run
replacement for ihnp4 and cbosgd.

We are also not cutting off netnews (e.g. Usenet.)  Our cutbacks all
involve email.  AT&T has a management committment to support netnews.

>  A major cornerstone of the future of USENET has to be services like uunet.
>  As AT&T goes, more and more backbones will be forced to follow, as folks
>  try to find other "free" services to feed their habits. And nobody, no
>  combination of backbones, can take up the slack for what AT&T's done. So
>  there will have to be a domino effect here. Uunet is going to be a
>  cornerstone, but I don't think it can do it alone. What will probably have
>  to happen is a new, "commercial" backbone of services like uunet that all
>  talk to each other and handle the connectivity of the network. No company
>  is going to be able to (much less willing to pay for) that anymore at the
>  current size.

There is, in fact, a major reshaping of the email world in the works.
The current system is cooperative - people worry about delivering mail
first, and getting it paid for isn't a concern.  The telephone network
was in a state similar to this in 1910 or so.  There were many disconnected
telephone systems, and some people offered gateways by subscribing to
more than one and holding phones up to each other.

With X.400 beginning to be offered, commercial mail services are springing
up.  These include ATTMAIL, Canada's Envoy, Britain's BTI, and so on.
There are others, not all are X.400: MCIMAIL, Easylink, Dialnet, etc.
(There are also services like UUNET which charge by connect time, and
which are nonprofit.  They don't seem to count.)

These services all charge the *sender* (except for COD mail) for each
message, and in general worry more about getting paid than about delivering
the message.  So they won't pass a message unless they know who to bill.
As long as the commercial services are small, there's a horrible barrier
between the two: nobody will admit to having a gateway because mail
through them would be charged to the gateway.  The commercial services
will only send mail to places they have signed service agreements with
each other.

As the commercial services grow (there is an assumption that the bulk of
the business users of email will sign up with commercial services instead
of rolling their own in a cooperative network, and that we technical folks
represent a tiny part of the market) the cooperative world will have more
and more interest in talking to customers of commercial services.  At
some point, the center of gravity will shift to the commercial services
and the cooperative folks will wind up hurting.  The money that's funding
the cooperative networks may dry up as the sugar daddies put it into
commercial systems instead.  Eventually we'll be at the level of HAM
radio: small and specialized, and unable to be self sufficient.  (A more
frightening thought is that we may become, indeed are becoming, like CB
radio: too much noise and too little signal.)

(By the way, the commercial world is serious about email, but doesn't
know what to do about netnews.  It doesn't seem interested in it.)

Over the next 5 or 10 years, the face of the world will change.  We must
either adapt to the change, or go away.  The commercial services point
out how much better off the world would be if everyone used the commercial
services - a maximum of 2 hops, support, reliablity, very fast cross
country delivery of email, access to the postal mail, telex, fax, etc.
They also point out how the cooperative networks are paying the costs,
often very high costs, but these costs are hidden as part of the usage
of people, machinery, phone lines, and so on.  On a nationwide basis,
there is certainly an economy of scale.

If you look at the telephone network (or the package delivery network)
you see there are common carriers, and there are local endpoints.
There is often no choice of local endpoint: you have one local phone
company, or one company shipping & receiving department.  But the
sender, who pays for the call/package, chooses a common carrier to pay
from several that are available.  The common carriers compete among
each other.

In our world, the cooperative networks are just another common carrier
(or group of common carriers.)  If it's cheaper to deliver your mail
yourself, you'll do it.  I'm not going to put a stamp on a note to my
5 year old, either, or on company internal mail.  PBX systems don't
charge for calls to another extension.

The commercial networks won't sign exchange agreements with cooperative
networks, however.  They want their money, and with random incoming
mail from UUCP, you don't know who to send the bill to.  In X.400 terms,
exchange agreements are for ADMDs (Adminstration Management Domains)
and not for PRMDs (Private Management Domains), and ARPA, UUCP, BITNET,
et al are PRMDs.

It may be necessary for the cooperative nets to become ADMDs, either
collectively or separately.  This could give them the clout and equality
they need to survive and still have universal service.  Each government
decides who is an ADMD, and as far as I know, the US government has not
yet stepped into the picture.

There is another model besides the "lots of common carriers, sender
chooses" model.  In that model, the commercial services are both
common carriers and endpoints.  To send mail to a user at a particular
endpoint, on a particular common carrier, you must pay that common
carrier.  Sender pays, but recipient chooses who the sender pays.
I'm told that this is becoming a de-facto standard, that the commercial
services put all their employees on their own service, as well as
their customers, and they can only be reached through their own service.
(In the case of two commercial carriers, there is a surcharge, not unlike
buying more stamps to send a letter to a foreign country.  The two
carriers split the postage.)  For a company that is new and their
employees were never reachable by email before, this might work.
But the eventual ramifications of this model for large companies,
such as IBM, which owns MCIMAIL, are clear.  It is for this reason
that I personally prefer the "whoever pays chooses who they buy
service from" model.

Anyway, some evolution of the cooperative networks is inevitible.
DOD wants out of the ARPANET business.  Seismo and AT&T want out
of the third-party-email-pass-through-for-free business.  Many users
of UUCP use UUCP email because no commercial service is available,
and will shift to the commercial services.  Those of us who are left
need to decide how we want to fit into this brave new world, and
adjust accordingly.

One thing we ought to be thinking about is making netnews more
independent of email.  Right now netnews depends on existing UUCP
email for lots of things: replies, moderated postings, control
messages, test messages, setup, error messages, etc.  (Can anyone
add to this list?)  As email as we know it crumbles, we need to
make netnews more self sufficient.

We also need to be thinking about fitting into the ADMD commercial
world somehow.

Comments?

	Mark

hack@bellboy.UUCP (Greg Hackney) (06/16/88)

In article <585@cbnews.ATT.COM> mark@stargate.COM (Mark Horton) writes:
 
> Mail into and out of AT&T through the "att" gateway will continue.
> Comments?

Only if it is from an AT&T site, or whose final destination
is an AT&T site (from what I understand). The question I must
ask myself is, do I want a connection bad enough to pay for the
phone costs of passing email to AT&T employees, for their benefit.

Defend the change as you will, but to me it marks the beginning
of a gradual exodus from the past committment to support USENET.

There's little doubt in my mind that this change in policy is
in conjunction with the advent of the commercial ATT Mail service.
There is also little doubt in my mind that such a decision would
be from management, rather than the technical folks that built
the network (and are the predominate users). It's sad to see a
shift in attitude, from one of being supportive of a educational
exchange of information, to a more purely commercial attitude.

I suspect it won't be too long before y'all cut back on netnews
feeds, allowing just enough to satisfy your internal needs.
But I can understand this, since yours IS a commercial enterprise.
(I wonder if they took a nation wide vote on rescinding Judge
Green's order, what the public will would be.)

My comment is, thanks for the memories. Hope you make a buck,
(although it won't be mine). And also, I think you underestimate
the grit of the "net" to heal itself of lost limbs.

vixie@palo-alto.DEC.COM (Paul Vixie) (06/17/88)

In article <1100@bellboy.UUCP> hack@bellboy.UUCP (Greg Hackney) writes:
# In article <585@cbnews.ATT.COM> mark@stargate.COM (Mark Horton) writes:
# > Mail into and out of AT&T through the "att" gateway will continue.
# > Comments?
# 
# Only if it is from an AT&T site, or whose final destination
# is an AT&T site (from what I understand). The question I must
# ask myself is, do I want a connection bad enough to pay for the
# phone costs of passing email to AT&T employees, for their benefit.

There are actually going to be three separate 'att' machines in three
different cities.  And mail sent x!att!y will work, even if x and y
are both non-AT&T sites.  I think AT&T's plan is basically to disallow
mail along x!att!another-att!yet-another-att!y, that is, they'd rather
not have their internal network used for non-AT&T traffic.

I don't blame them.  Most companies have a well-connected gateway which
forwards vast multitudes of local and some long distance traffic --
between the gateway's non-company neighbors -- in addition to being
the point of contact between all internal sites and all external sites.

I don't think AT&T will only speak to their neighbors to send and receive
AT&T traffic.  They could do this, but it doesn't seem likely to me.
It's more likely that 'att' will still handle thousands of messages
per day of non-AT&T traffic, picking them up from direct UUCP neighbors
and sending them to others.

Does this bug you?  I'm sure all of "att"'s direct UUCP neighbors will
be well served by "att" as a local hub, well enough served that "att"'s
use of their neighbors to forward AT&T's internal traffic will not be
a bother.

Again, AT&T seems to be saying: use our machine, but not our internal
network.  And given the business they're in, that makes great sense to me.

Disclaimer: I'm not speaking for Digital.
-- 
Paul Vixie
Digital Equipment Corporation	Work:  vixie@dec.com	Play:  paul@vixie.UUCP
Western Research Laboratory	 uunet!decwrl!vixie	   uunet!vixie!paul
Palo Alto, California, USA	  +1 415 853 6600	   +1 415 864 7013

haugj@pigs.UUCP (The Beach Bum) (06/18/88)

In article <3095@palo-alto.DEC.COM>, vixie@palo-alto.DEC.COM (Paul Vixie) writes:
> I don't think AT&T will only speak to their neighbors to send and receive
> AT&T traffic.  They could do this, but it doesn't seem likely to me.
> It's more likely that 'att' will still handle thousands of messages
> per day of non-AT&T traffic, picking them up from direct UUCP neighbors
> and sending them to others.

This is wrong given the information that I have.  My understanding is that
if the `From' machine and the `To' machine are both not part of either
att.com or, if the `To' machine is not on the ARPAnet, that the letter
will be dropped.  It is also my understanding that AT&T has already put
this into effect on several systems.

For those who may wish to defend AT&T, they are the ONLY company which I
am aware of which will not transport third part mail, even if the third
party is a local call to an AT&T gateway.  Which means, Illinios and
New Jersey are both going to suffer.

- John.
-- 
 The Beach Bum                                 Big "D" Home for Wayward Hackers
 UUCP: ...!killer!rpp386!jfh                          jfh@rpp386.uucp :SMAILERS

 "You are in a twisty little maze of UUCP connections, all alike" -- fortune

webber@porthos.rutgers.edu (Bob Webber) (06/18/88)

In article <3095@palo-alto.DEC.COM>, vixie@palo-alto.DEC.COM (Paul Vixie) writes:
> ...
> There are actually going to be three separate 'att' machines in three
> different cities.  And mail sent x!att!y will work, even if x and y
> are both non-AT&T sites.  I think AT&T's plan is basically to disallow
> mail along x!att!another-att!yet-another-att!y, that is, they'd rather
> not have their internal network used for non-AT&T traffic.
> ...
> I don't think AT&T will only speak to their neighbors to send and receive
> AT&T traffic.  They could do this, but it doesn't seem likely to me.
> It's more likely that 'att' will still handle thousands of messages
> per day of non-AT&T traffic, picking them up from direct UUCP neighbors
> and sending them to others.

You know, that made so much sense that I went back and reread the
news.announce.important announcement from AT&T to see if everyone else
(myself included) had been misreading it.

<Path: aramis.rutgers.edu!rutgers!mtunx!att!cblpf!ihnp4!attnews
<From: attnews@ihnp4.att.com (Harold Jackson)
<Sender: mark@cblpf.ATT.COM
<...
<This overload has to be rectified.  Management has decided to
<discontinue passing third party email through AT&T machines. ...
<This cutoff applies to ihnp4, cbosgd, and att.  It will probably affect
<other AT&T machines within a year.
<IF YOU LIST IHNP4, CBOSGD, OR ATT AS SMART-HOST, OR IN YOUR
</USR/LIB/NEWS/PATHS FILE, YOU SHOULD CHANGE IT *NOW* TO SOME OTHER
<MACHINE.
<While we are phasing out third party pass-through as much as possible,
<we will continue to accept email for AT&T machines, and to
<pass outgoing email from AT&T to other systems. 

While the rest could probably be stretched to your scenerio, the stuff
in all caps seems to indicate that nonatt-oneatt-nonatt is not going
to be permitted.

------ BOB (webber@athos.rutgers.edu ; rutgers!athos.rutgers.edu!webber)

casey@admin.cognet.ucla.edu (Casey Leedom) (06/22/88)

In article <3095@palo-alto.DEC.COM> vixie@palo-alto.DEC.COM (Paul Vixie) writes:
> In article <1100@bellboy.UUCP> hack@bellboy.UUCP (Greg Hackney) writes:
> # In article <585@cbnews.ATT.COM> mark@stargate.COM (Mark Horton) writes:
> # > Mail into and out of AT&T through the "att" gateway will continue.
> # > Comments?
> # 
> # Only if it is from an AT&T site, or whose final destination
> # is an AT&T site (from what I understand).
> 
> There are actually going to be three separate 'att' machines in three
> different cities.  And mail sent x!att!y will work, even if x and y
> are both non-AT&T sites.  I think AT&T's plan is basically to disallow
> mail along x!att!another-att!yet-another-att!y, that is, they'd rather
> not have their internal network used for non-AT&T traffic.
> 
> I don't blame them.  Most companies have a well-connected gateway which
> forwards vast multitudes of local and some long distance traffic --
> between the gateway's non-company neighbors -- in addition to being
> the point of contact between all internal sites and all external sites.
> ... Again, AT&T seems to be saying: use our machine, but not our internal
> network.  And given the business they're in, that makes great sense to me.

  Someone quite correctly pointed out that Mark's letter seems to imply
that x!att!y will be bounced, but it doesn't really matter one way or the
other.

  Even if x!att!y will be allowed, Paul's argument falls down because it
states that it's perfectly ok for AT&T to say ``We're not going to let
people to use our internal network links.'', and [implicitly] that it's
ok for AT&T to use USENET's internal network links (USENET is of course
not owned by a single organization, but it is an informal COOP).

  Why should a user on site "att" be able to send messages to X!Y!Z using
the network links between X and Y, and Y and Z?  X, Y, and Z may belong
to the same organization or separate organizations, but they may all be
considered to belong to the virtual USENET organization.

  According to Paul's argument, the AT&T user shouldn't be allowed to do
this because that user would be using internal network links within an
organization (i.e. USENET).  As it turns out, cooperative message
exchange is the entire business of USENET and so X!Y!Z *is* ok.  What
AT&T is proposing is that they drop most of their message carrying duties
in the cooperative message exchange, but still retain full membership
rights with regards to their own message loading.

  Again, if AT&T were a poor site without much money or traffic, network
charity would be the name of the game.  But AT&T is not deserving of such
charity.

  I don't want to knock the job that AT&T UUCP sites and their
administrators have done, but it was they who took on those untold number
of links.  If the load is too heavy, they should either cut back on the
number of links and still remain full partners in USENET, or, if they
want to be a leaf node, they should subscribe to uunet just like anyone
else.

Casey

vixie@palo-alto.DEC.COM (Paul Vixie) (06/25/88)

In article <13758@shemp.CS.UCLA.EDU> casey@admin.UUCP (Casey Leedom) writes:
#   Even if x!att!y will be allowed, Paul's argument falls down because it
# states that it's perfectly ok for AT&T to say ``We're not going to let
# people to use our internal network links.'', and [implicitly] that it's
# ok for AT&T to use USENET's internal network links (USENET is of course
# not owned by a single organization, but it is an informal COOP).
# 
#   Why should a user on site "att" be able to send messages to X!Y!Z using
# the network links between X and Y, and Y and Z?  X, Y, and Z may belong
# to the same organization or separate organizations, but they may all be
# considered to belong to the virtual USENET organization.

So does AT&T, though, belong to the virtual USENET organization.

#   Again, if AT&T were a poor site without much money or traffic, network
# charity would be the name of the game.  But AT&T is not deserving of such
# charity.

Charity?  Hey, but there are a lot of very smart people inside AT&T that I
want to exchange mail messages and news articles with.  Sure, AT&T brought
us System V, but there are other parts of the company that have perfectly
decent and fun people :-).

It's not charity.  The only thing anybody has ever had a right to ask of any
USENET site is that they be willing to forward mail and/or news to other sites;
AT&T is going to keep doing this, probably on a large scale, but only from one
point inside their organization.

There's no way anyone can require that AT&T provide UUCP access to all the
machines on their internal net -- the administrative headache alone is more
than they can afford to WASTE.  They are on the net for their benefit, and
if they drop out because of rudeness on our part, we will incidentally be
much worse off for it.

#   I don't want to knock the job that AT&T UUCP sites and their
# administrators have done, but it was they who took on those untold number
# of links.  If the load is too heavy, they should either cut back on the
# number of links and still remain full partners in USENET, or, if they
# want to be a leaf node, they should subscribe to uunet just like anyone
# else.

Um, yeah, they can subscribe to uunet.  Uunet's customers would have to
touch all that outgoing mail, though, and the customers pay for it.  I.e.,
mail from att to uunet!foo!bar is going to be paid for by "att" and "foo".

What would you prefer -- that AT&T maintain a machine on which they forward
a lot of local traffic (between their direct neighbors), or that they
disappear altogether?
-- 
Paul Vixie
Digital Equipment Corporation	Work:  vixie@dec.com	Play:  paul@vixie.UUCP
Western Research Laboratory	 uunet!decwrl!vixie	   uunet!vixie!paul
Palo Alto, California, USA	  +1 415 853 6600	   +1 415 864 7013