Xref: utzoo soc.culture.british:11427 comp.protocols.tcp-ip:16415
Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!wuarchive!uunet!munnari.oz.au!bunyip.cc.uq.oz.au!brolga!ggm
From: ggm@brolga.cc.uq.oz.au (George Michaelson)
Newsgroups: soc.culture.british,comp.protocols.tcp-ip
Subject: Re: IP in the UK (was Re: Fingering the English)
Message-ID: <1991Jun6.233329.5232@brolga.cc.uq.oz.au>
Date: 6 Jun 91 23:33:29 GMT
References: <7957@ecs.soton.ac.uk> <5280@syma.sussex.ac.uk> <1991Jun06.062415.1848@lut.ac.uk>
Organization: Prentice Computer Centre, The University of Queensland, Australia.
Lines: 75

I suspect that my lack of knowledge of the internal architecture of the
IP/X25 connections is the source of  some confusion for me. 

-I'm assuming that you are doing IP tunnels to one common endpoint
which is feeding into the Internet, and thus there is no "logical"
common backbone beyond a star-hub (a cisco in ulcc or ucl?).  

I'm also assuming that down the track this will not scale well, and
some degree of distribution of the load will prove neccessary. 

-That means that IP routing boxes will be inter-connecting to form a
backbone, and there WILL be some kind of shared usage of trunk lines.

Question: would *THAT* traffic be done using X.25 tunnels as well?

based on those assumptions you can now go and destroy all of what follows!

> ccrth@lut.ac.uk (Rob Thirlby) writes:

>The decision to use IP tunelling rather than multiplexing our access
>lines to janet reflects the very variable and unpredictable demand from
>sites and the wide disparity of bandwidth of the access lines.  Also
>many of us are hoping to use tunelling facilities in our existing kit
>such as netcomm switches/ECB rather than buy unnecessary routers.


I have some doubt thats going to prove "optimal" in the longer term, but
I also wonder if how individual campii connect into a backbone should
dictate how the backbone allocates its bandwidth? I wouldn't want the NSF
to allocate 48k trunks simply because AARN is only using 48k for the drops!

I think this depends on what you think the long-term implications of a
wide-band connectionless network are. What sort of dataflow down the
shared links do you expect?  If you tunnel each campus connection over
X.25, will some pool of JANET backbone X.25 routers do the right thing
with all the packets going to the US and onward?

I'd expect that the use of long-haul links will look very much the same
for you as they do for everybody else: NNTP, SMTP and FTP traffic will
be semi-continuous, with a layering of telnet and other small packet
traffic during "normal working hours". -That means that shared
resources will be blasted with big packets all flowing in pretty much
the same direction:

	outer-lying places to FTP archives
	outer-lying places to newsfeeds
	outer-lying places to mailhubs
	everywhere to the International fatpipes.

Surely having boxes which understand the IP addressing and co-elesce all
this into more optimal flow down HDLC is better than maintaining 1001 separate
X.25 tunnels all going to the same place?

>We are getting there.  The pilot scheme is being persued with vigour and
>many sites such as mine will ensure that the pressure to offer a 
>service is kept up!  After all isnt it said in religious circles
>that late converts make the most devout members?

Point taken. I must lay claim to fitting that pattern, I doubt if
anybody in the UK would recall me being exactly the most pro-TCP of
people. However like the small boy in the "Bubbles" advert:

	since using their soap, I have used no other.

I really hope the project continues to flourish. If I'm provoking any
anger please accept my apologies, it's a reflection of the long long time
this has been coming. -Roll on '92!

-George

-- 
                         George Michaelson
G.Michaelson@cc.uq.oz.au The Prentice Centre      | There's no  market for
                         University of Queensland | hippos in Philadelphia
Phone: +61 7 365 4079    QLD Australia 4072       |          -Bertold Brecht