🚀 go-pugleaf

In article <1989May14.165359.21028@telly.on.ca> evan@telly.on.ca (Evan Leibovitch) writes:
>Upon redoing my paths file recently, I noticed that pathalias now considers
>the fastest trip from here to uunet to be
>
>telly!utzoo!utgpu!utai!watmath!uunet
>
>rather than what it used to be, telly!attcan!uunet

I also noticed this yesterday, so I tried mail to:
	mason%tmsoft@uunet.uu.net
 and	mnetor!uunet!mason%tmsoft	(my previous best path)
to test the time it took via each path.

The one via mnetor worked fine (although it didn't go out of Toronto
until midnight (NOTE: I'm definitely NOT complaining) but it got
turned around immediately at uunet & was back here within 30 minutes).

The one via watmath translated into utcsri!utai!watmath!uunet!... but
it bounced off utgpu, yes gpu!  It seems that watmath's internet path
actually goes via utgpu (though why I don't know!) but it mangled the
bounce path to such an extent that gpu couldn't make sense of it and I
got some nice mail from Postmaster@gpu.  Needless to say, I re-ran
pathalias with -dwatmath!uunet.  Everything is now fine.

>This brought on a couple of curiosities:
>
>1) What's changed in local connectivity to bring this about? Does watmath
>   really talk to uunet every 1/2 hour?
See above.  Sort-of is my evaluation of this.

>2) I thought UofT also talks to uunet. If the waterloo connection is so
>   good, why bother?
My understanding is that UofT does not want to be the local mail hub
any more than they already are.  Note that the waterloo internet
connection PHYSICALLY (if not logically, see above) goes through UofT.
Moreover mail from Toronto will follow some UofT path to Waterloo,
then back to UofT to head on to uunet.  Not very efficient.  I thought
Rayan was going to control the costs so Waterloo would not look like
the path of choice from Toronto.  If anyone is going to advertise an
internet connection, it would make more sense for it to be UofT, but I
gather the politics of the situation prevent it.

>4) Can anyone confirm what I've heard as rumour, that Telebit
>   LD links for news feeds are actually cheaper than X.25?
Talk to ron@mnetor.

	../Dave

In article <1989May14.165359.21028@telly.on.ca> evan@telly.on.ca (Evan Leibovitch) writes:
# Upon redoing my paths file recently, I noticed that pathalias now considers
# the fastest trip from here to uunet to be
# 
# telly!utzoo!utgpu!utai!watmath!uunet
# 
# rather than what it used to be, telly!attcan!uunet

This does seem rather strange indeed.  The path is definitly not what
it should be.  As far as I can tell watmath calls uunet using the UofT
Internet connection. The utgpu <-> watmath link is over ONET.  The message
travels to UofT, thence to waterloo over ONET, back over ONET to UofT, and then
down to uunet.  The shortest path would be telly!utgpu!uunet but the appropriate
path should be telly!attcan!uunet -- utai is no closer to the Internet than
any other host at UofT.

My concern, however, lies elsewhere.  I'm sure the folks at Waterloo didn't
intend to advertise the NSFnet connection at UofT as the default path down
to uunet.  AT&T Canada have graciously offered to handle email traffic to
uunet.  It is not appropriate to use the NSFnet link at UofT for this.

-- 
Tom Molnar
Unix Systems Group, University of Toronto Computing Services.

In article <1989May14.165359.21028@telly.on.ca> evan@telly.on.ca (Evan Leibovitch) writes:
>4) Can anyone confirm what I've heard as rumour, that Telebit
>   LD links for news feeds are actually cheaper than X.25?

Rick Adams of UUNET has pretty much said as much.  It may depend somewhat
on details, but the fact is that X.25 is an expensive way to ship bulk
data:  the connect charge does not look that high, but the per-packet or
per-byte charges get nasty for bulk transmission, and the throughput is
not that good.  A Trailblazer link after midnight is a pretty cheap data
pipeline.
-- 
Subversion, n:  a superset     |     Henry Spencer at U of Toronto Zoology
of a subset.    --J.J. Horning | uunet!attcan!utzoo!henry henry@zoo.toronto.edu

In article <1989May14.191801.18773@tmsoft.uucp> mason@tmsoft.UUCP (Dave Mason) writes:

# >2) I thought UofT also talks to uunet. If the waterloo connection is so
# >   good, why bother?
# My understanding is that UofT does not want to be the local mail hub
# any more than they already are.

Correct.  I'm sure you can appreciate why.  On average, utgpu handles
~3000 message a day.  One evening I decided to poke through the logs to
see what the traffic pattern was.   I was surprised to see that for that
particular day, only 20% of the traffic was either to or from a UofT mailbox,
the rest was gatewayed mail.  Naturally that covered only a single 24 hour
period, but the discovery was interesting nonetheless.

Tom Molnar
-- 
Tom Molnar
Unix Systems Group, University of Toronto Computing Services.

>In article <1989May14.165359.21028@telly.on.ca> evan@telly.on.ca (Evan Leibovitch) writes:
>4) Can anyone confirm what I've heard as rumour, that Telebit
>   LD links for news feeds are actually cheaper than X.25?

The arithmetic is pretty simple, between two points in Ontario Datapac costs
between $3 and $6 per Kilo-packet or 256K, direct dial long distance rarely 
exceeds $.30/minute. I won't bore anyone with the actual calculations, the 
resulting costs end up looking something like:

	Datapac: 1.1 - 2.3 cents/Kilobyte
	TeleBit: 0.6       cents/Kilobyte

The Telebit costs go down even further if you use then after 11PM, the other
thing you can do with trailblazers that is not recommended with lesser modems
is run them on WATS lines. 

The CRTC announced this morning that Bell has been ordered to stop subsidizing
data services from regular telephone service income, as a result prices will be
going up June 5th for DataPac, Inet2000, Envoy 100, Dataroute and more.

-- 
Norman Soley - The Communications Guy - Ontario Ministry of the Environment
Until the next maps go out:	moegate!soley@ontenv.UUCP 
if you roll your own: 	uunet!{attcan!ncrcan|mnetor!ontmoh}!ontenv!moegate!soley
I'd like to try golf but I just can't bring myself to buy a pair of plaid pants

In article <1989May15.234059.29449@gpu.utcs.utoronto.ca> molnar@gpu.utcs.UUCP (Tom Molnar) writes:
>I was surprised to see that for that
>particular day, only 20% of the traffic was either to or from a UofT mailbox,
>the rest was gatewayed mail.

It would be illuminating to see how much of that traffic was entirely local
and how much involved the Internet or long distance. I know that I often
route mail from here through UofT to other local sites (utgpu!geac,
utcsri!alias) because it ends up being faster.

Maybe you do your job too well? :-)

-- 
C. Harald Koch		NTT Systems, Inc., Toronto, ON Canada
chk@dretor.dciem.dnd.ca, chk@gpu.utcs.toronto.edu, chk@chkent.UUCP

In article <1989May15.234059.29449@gpu.utcs.utoronto.ca> molnar@gpu.utcs.UUCP (Tom Molnar) writes:
># My understanding is that UofT does not want to be the local mail hub
># any more than they already are.
>
>Correct.  I'm sure you can appreciate why...

Incidentally, if anyone has ever wondered why utzoo adamantly refuses to
automatically reroute mail, it's the same reason:  we don't want to
encourage people to use us as a router.
-- 
Subversion, n:  a superset     |     Henry Spencer at U of Toronto Zoology
of a subset.    --J.J. Horning | uunet!attcan!utzoo!henry henry@zoo.toronto.edu

In article <1989May15.234059.29449@gpu.utcs.utoronto.ca> molnar@gpu.utcs.UUCP (Tom Molnar) writes:
>In article <1989May14.191801.18773@tmsoft.uucp> mason@tmsoft.UUCP (Dave Mason) writes:
>
># >2) I thought UofT also talks to uunet. If the waterloo connection is so
># >   good, why bother?
># My understanding is that UofT does not want to be the local mail hub
># any more than they already are.
>
>Correct.  I'm sure you can appreciate why.  On average, utgpu handles
>~3000 message a day.  One evening I decided to poke through the logs to
>see what the traffic pattern was.   I was surprised to see that for that
>particular day, only 20% of the traffic was either to or from a UofT mailbox,
>the rest was gatewayed mail.  Naturally that covered only a single 24 hour
>period, but the discovery was interesting nonetheless.

Out of curiousity, I went through the u.can.on.? files looking for
uunet connectivity:

	jtsv16	uunet(HOURLY/2)
	dptcdc	uunet(HOURLY/2)
	bdofed  uunet(DEMAND)
	attcan	uunet(EVENING)
	mitel	uunet(DEMAND)
	mnetor	uunet(DEMAND+HIGH)
	pantor	uunet(DEMAND)
	paqs1	uunet(DEMAND)
	simware	uunet(DEMAND)
	watmath	uunet(DEDICATED)
	utai	uunet(HOURLY/2)

Assigning numeric costs to some of the symbols, we get:

	DEDICATED 200
	DEMAND	300
	HOURLY	500
	EVENING	1800

This has always confused me, since I thought attcan had graciously
expressed willingness to forward mail to and from uunet, and attcan
has the worst cost in the list!  We've always had watmath declared
dead, so we didn't get caught by their change to DEDICATED.  Considering
how well U of T is connected locally, is it any wonder that U of T ends
up forwarding everything?  Without extra hacking around with a local path 
file, almost every entry in our path file has "utai!uunet" in it.

Incidentally, most of the DEMAND entries were inserted by uunet personnel.

Anyways, I'm going to reinsert

	attcan	uunet(0)

into my path.local file and have done with it.
-- 
---------------------------------
Chris Lewis, R.H. Lathwell & Associates: Elegant Communications Inc.
UUCP: {uunet!mnetor, utcsri!utzoo}!lsuc!gate!eci386!clewis
Phone: (416)-595-5425

In article <302@moegate.UUCP> soley@moegate.UUCP (Norman S. Soley) writes:
# 	Datapac: 1.1 - 2.3 cents/Kilobyte
# 	TeleBit: 0.6       cents/Kilobyte

This is for Datapac-3101 service (i.e. they do PAD work for you), not
Datapac-3000 (you do your own PAD).  For the latter, rates are about 0.60$/kp
within Ontario (location-sensitive), which changes the picture somewhat.
However, to the U.S. there is much more of a balance, at that point the
precise billing schemes on phone lines become very important -- for example
uunet charges for 3 minutes minimum per call.  Also remember that it is
more expensive to have an empty call with a Telebit (1-2 minutes) than with
UUCP/X.25 (50 packets?).  It gets complicated.  I think that with 1 minute
minimum and night rates, TB's win out.  Otherwise it gets pretty fuzzy which
is better.  Oh yes, remember to factor in the monthly charges if you do
your own PAD work, you need nontrivial volume for them to disappear in the
wash.

In article <1989May16.175732.4280@eci386.uucp> clewis@eci386.UUCP (Chris Lewis) writes:
	....We've always had watmath declared
	dead, so we didn't get caught by their change to DEDICATED.

FYI: the watmath entry had the uunet connection marked as EVENING+LOW  (2005)
( by me ) a couple of months ago and at that time I noticed that pathalias
was generating weights of DEDICATED (100) no matter what I did.  I put a
comment line in our map entry saying so.  I guess pathalias was working
ok everywhere else. ( yes, I can reproduce it ).  I've been too busy to look
into it.  It was also producing alot of 'uunet!utgpu!...' when it thought
we had the same weight to uunet and utgpu.
I just since found out that our map entry had been changed, declaring uunet
as DEDICATED.  And thats when everybody started noticing it.  sorry. 

I've just changed our map entry to what Rayan suggested, pathalias seems
to be generating correct weights now, and I'll send in the new map entry.
	-rj