Internet routing Europe -> USA -> Europe...
wolf paul
)While trying to find whether we (in France, Europe) could reach a site
)in Germany (Europe), I got the following route from traceroute:
)
)I.e.: Paris -> South France -> NJ -> PE -> Ithaca (upst. NY)->
) Syracuse (upst. NY) -> NYC, NY -> ? -> Deutschland
)
>Why this contorted route? Is it cost-effective?
Well, I guess it has to do with the cost of leased lines used for
internet connections. Since the bulk of internet activity takes place
in Europe, most European countries have more direct links to
sites in the US than they have to sites in other European countries.
Since these leased lines are not really charged by volume but rather
have a fixed monthly charge regardless of traffic, it probably does not
affect the cost a whole lot.
The Austrian branch of EUnet will shortly be connected to the Internet
by a leased line from tuvie to mcsun; our organization may also get a
leased line to tuvie, thus any internet connections from here to
France will run via Holland. Mcsun is connected to the U.S.; unless
there is a direct connection from mcsun to some French site you talk
to directly, a connection from here to you would also run via the
U.S.
--
Wolf N. Paul, IIASA, A - 2361 Laxenburg, Austria, Europe
PHONE: +43-2236-71521-465 FAX: +43-2236-71313 UUCP: uunet!iiasa.at!wnp
INTERNET: wnp%iias...@uunet.uu.net BITNET: tuvie!iiasa!w...@awiuni01.BITNET
* * * * Kurt Waldheim for President (of Mars, of course!) * * * *
Michel Fingerhut
> Since these leased lines are not really charged by volume but rather
> have a fixed monthly charge regardless of traffic, it probably does not
> affect the cost a whole lot.
Well, it is not true insofar as the end user (which I am). As it
appears, the French backbone will charge us Internet mail by volume.
It has to do, apparently, with the cost of some line whose cost is
fixed, but which they intend to share "equally well" among users -- ie
the more you use it the more you pay for it.
He adds:
> The Austrian branch of EUnet will shortly be connected to the Internet
> by a leased line from tuvie to mcsun [in Holland]
Too bad. Although there is a line from France to Northern Yurop, there is
no connection for academic sites from France to there. There is no rerouting
through anywhere else, either. So this means that THERE IS NO INTERNET
CONNECTION between any sites connected to mcsun and France. Mail goes around,
I hope, but since I did not get to send any yet who knows.
I suppose that when the various backbones sort their differences, we'll have
to pay so as to get to Northern Y. I hope *they* will have to pay to get to
us, but I suspect there is more interest in the connection from us out than
the converse. So we loose.. Unless we give up on Northern Y. altogether
and look as always in the US for software. So much for CEE, connectedness
and other grandiose ideas.
Eric Thomas SUNET
problem. France is basically 5 years behind most other "rich" european
countries in terms of networking. There is a serious lack of TCP/IP
connectivity as you have mentioned. There is also a serious lack of decent
gateways between UUCP, the internet, BITNET, etc. Just yesterday I received
a message from some french UUCP site, through a gateway at ENS Lyon. The
'Return-Path:' field contained just the login name of the poster on the UUCP
machine, and the 'From:' field contained some UUCP routing information and no
host name. Needless to say it was impossible to reply to this message, apart
from the fact that RFC822 mailers don't know what to do with UUCP routing the
syntax of the field was simply invalid. I complained, but I doubt anything will
happen, meanwhile french users "in the know" make use of the CERN gateway,
because "it works", and french politicians are happy that the particular site
they are in charge of is getting more influence and that they have managed to
get some prestigious position in RARE Working Group so-and-so, and who cares
about the rest? :-)
Eric
PS: Sweden is not part of the EEC, and I am not precisely enthusiastic about
the way the EEC spends its networking money, but that is another story...
Piet Beertema
>>Why this contorted route? Is it cost-effective?
>
>Well, I guess it has to do with the cost of leased lines used
>for internet connections.
it's just a small technical problem (a gateway not
announcing a route) that can be solved quickly.
--
Piet Beertema, CWI, Amsterdam (pi...@cwi.nl)
Michel Fingerhut
backbone effort to supply tcp-ip connectivity. I am all for it. I also
know that their resources are very limited, so I can't blame them for the
problems. Besides, I don't care about blame, but believe in pointing at
the problems, hoping they will resolve sometime somehow.
One of the "objective" reasons, I believe, is that Unix and related "culture"
made so little headway in France when compared with other European countries.
I have some ideas about what the reason is. But forget it, it's the past.
I do disagree with Eric Thomas' denying it being a EEC or European problem,
because yes, it is one too: (i) if there is no connectivity for academic
sites in France with all of northern Europe, it is because of "disagreements"
between France and Holland. Both ends (or sides) of the connection.
I fear that too small a number of sites is actually interested in full Internet
connectivity, and so the financial onus will be much greater for those who
want it. We being a non-profit educational and research center, I wonder
whether or not we will be able to carry our share of the burden. We are
not told what it may amount to in the coming years. Unfortunately, money
does not grow on the trees in my neighborhood.
I think it is a cultural problem. And that will be the hardest to change,
I'm afraid.
Michael Fingerhut
Eric Thomas SUNET
a T1 line between Cornell University and CERN, for instance, and the thing
burning the biggest amount of bandwidth on this line is BITNET traffic between
two IBM machines. Likewise one of the major "users" of the line between
Stockholm and Amsterdam is EARN traffic, again between 2 IBM machines; thanks
to TCP/IP we were able to get rid of the old 9600bps leased line that was
dedicated to EARN and save money (and with more money comes more bandwidth or
more lines). In Stockholm we have a high-speed microwave link between several
institutes, and most of these sites have VMS VAXes; they run both DECNET over
IP and "native" TCP/IP on the machines in question. Let's face it, TCP/IP is
the industry standard, it works very well and it's available on all the
operating systems, even on MVS! You don't need a Unix culture to want TCP/IP
(I for one have never been accused of having a Unix culture :-) :-) ).
I think the problem with France is that everybody wants to do his own little
backbone with his favourite protocol instead of putting all the money in a
common IP backbone as we do in Nordunet for instance. How many EARN links are
running on top of TCP/IP in France? The last time I checked, there wasn't a
single one. This means each and every EARN link requires a dedicated line,
and sites which want EARN+IP are paying for 2 lines (and then complaining they
don't have money for more bandwidth). On the other hand when there basically
does not exist any backbone it is difficult to avoid long-distance dedicated
lines.
Eric
Michel Fingerhut
>It has nothing to do with costs, politics or such; it's just a small
>technical problem (a gateway not announcing a route) that can be solved
>quickly.
Well the "official" response of the French backbone is that the (non) connecti-
vity of (academic sites in) France to Nordunet IS a problem of politics and
costs. Between the Netherlands and France.
Roland Dirlewanger
European audience. Well, for the defence of French networking:
In article, <...> m...@ircam.ircam.fr (Michel Fingerhut) writes:
> Too bad. Although there is a line from France to Northern Yurop, there is
> no connection for academic sites from France to there. There is no rerouting
> through anywhere else, either. So this means that THERE IS NO INTERNET
> CONNECTION between any sites connected to mcsun and France. Mail goes
around,
> I hope, but since I did not get to send any yet who knows.
Situation in France is as follow: there is an NFSnet line connected from
INRIA
(Sophia-Antipolis) to the JVNC gateway in the US. The deal between Inria
and
NFS is, as far as I know, that every academic site of Southern Europe
may use
this link for free as soon as it reaches INRIA-Sophia at its own costs.
Ircam
is connected to INRIA-Rocquencourt then uses INRIA's private network
from
Rocquencourt to Sophia and reaches the US-Internet. So far so good.
Concerning Northern-Europe, there's no one who pays for the INRIA -> CWI
link
(about 200 000 FF a year fixed costs + 100 000 FF for a Cisco box +
maintenance
for the whole stuff) so the cost has to be shared by all its users.
Chorus agreed to share the costs for this link. So, I can use this link
to reach
US-Internet as well as European hosts (InterEUnet). I claim that
Internet
connectivity *exists* from France to Europe but, alas it's not free of
charge !
In article <20...@sunic.sunet.se>, er...@sunic.sunet.se (Eric Thomas
SUNET) writes:
|> I don't want to sound mean, but this isn't an EEC problem, this is a
french
|> problem. France is basically 5 years behind most other "rich"
european
|> countries in terms of networking. There is a serious lack of TCP/IP
|> connectivity as you have mentioned. There is also a serious lack of
decent
|> gateways between UUCP, the internet, BITNET, etc. Just yesterday I
received
|> a message from some french UUCP site, through a gateway at ENS Lyon.
The
|> 'Return-Path:' field contained just the login name of the poster on
the UUCP
|> machine, and the 'From:' field contained some UUCP routing
information and no
|> host name. Needless to say it was impossible to reply to this
message, apart
|> from the fact that RFC822 mailers don't know what to do with UUCP
routing the
|> syntax of the field was simply invalid.
Who's to blame ? The administration of the French network ? I don't
think so.
If the guy who sent you this E-mail had posted it by PTT mail with a
wrong return
address, could you *really* blame the French PTT for that ?
If any decent gateway receives an unqualified address from some host,
what should
it do ? Reject the mail, qualify it with the wrong domain name ? None of
these
solutions are corect, the problem must be treated at the source host.
--
Roland Dirlewanger E-mail: r...@chorus.fr
Chorus systemes Phone: +33 1 30 57 00 22
6, Avenue Gustave Eiffel Fax: +33 1 30 57 00 66
F-78182 St Quentin en Yvelines Cedex
Simon Poole
......
>US-Internet as well as European hosts (InterEUnet). I claim that
>Internet
>connectivity *exists* from France to Europe but, alas it's not free of
>charge !
What's "alas" about this? Either you pay for your connectivity yourself
(and have a say in how its acheived), or you just have to take what ever
is provided by your government. I don't see anything for you to complain
about. Or are free rides in networking a fundamental human right?
--
------------------------------------------------------------------------
Simon Poole
po...@verw.switch.ch / po...@chx400.switch.ch / mcsun!chx400!poole
------------------------------------------------------------------------
Eric Thomas SUNET
>If the guy who sent you this E-mail had posted it by PTT mail with a wrong
>return address, could you *really* blame the French PTT for that ?
>
>If any decent gateway receives an unqualified address from some host, what
>should it do ? Reject the mail, qualify it with the wrong domain name ?
>None of these solutions are corect, the problem must be treated at the source
>host.
This is not the problem I was complaining about. I got a 'From:' field saying
'From: host1!host2.fr!user'. I suppose that the actual uucp path was
something like 'host1!host2!user', and the gateway added a '.fr' to host2 for
some reason and then decided this is good enough and sent it to me. This is not
a failure of the uccp software at the sender's node, this is a brain-damaged
gateway. A gateway that purports to generate RFC822 headers should never give
you a header which does not respect the RFC822 syntax, period. It is obviously
the task of the gateway to translate 'host1!host2!user' to, say, 'us...@host2.fr'
(if such a hostname can be recognized via some MX reference) or at least
'host1!host2!us...@gateway.hostname.fr'. If you place a mail order for 12l of
wine and the postman opens your mailbox and pours 12l of wine into it, you will
probably be upset at not having received your wine in a proper packaging :-)
Eric
Daniel Karrenberg
A repetition of what Yves said for those who only read short articles:
1) Unfortunately we don't have a government funded European IP
infrastructure.
2) Because of 1) several organisations are funding IP links for their
paying members.
3) If you open those links to all potential customers immediately they
will not even notice they get a free ride paid by someone else. So
it is blocked until they notice.
Daniel
--
Daniel Karrenberg Future Net: <d...@cwi.nl>
CWI, Amsterdam Oldie Net: mcsun!dfk
The Netherlands Because It's There Net: DFK@MCVAX
Yves Devillers
>
> >>Why this contorted route? Is it cost-effective?
> >
> >Well, I guess it has to do with the cost of leased lines used
> >for internet connections.
> It has nothing to do with costs, politics or such;
> it's just a small technical problem (a gateway not
> announcing a route) that can be solved quickly.
>
-->You are *all* right. The problem is both technical and political
It's technical since the route to destination in Europe has to be received
from mcsun (provided it has to exists!) since mcsun is our next hop to
destination. Mcsun, according to an agreement we have inside RIPE, blocks
any propagation of that route to us (see below for rationale)
The fact that the route used passes via the US is an IP technical issue.
Routing protocols try to minimize the number of routes they propagate
rather than pass information for the 2500 to 4500 known Internet network
in every updates (every 90 seconds or so) in order to save on
bandwidth (existing 19.2 or 64kb/s commonly used in Europe is not
unlimited bandwidht)
Thus most (all) routers propagate few (no more than few hundreds) explicit
routes and rely on one or more default routes directing to the most populated
part of the IP world (read: the USA).
Consider the example of a sender leaf node site S sending a packet to
destination D with S being connected to major router R
If no route exist - on router R - for destination D (which is
intentionally the case if router R don't want to forward packet for
destination D or for sender S. See below for reasons of this policy routing)
router R will forward the packet over its default route (read: to the USA)
It might well be that the NSF backbone knows of a route to both D and S (they
have long ago solved most of their infrastructure costs through huge federal
money and "huger" regional and local money so they can safely consider D and
S as peer to which they render equivalent services whatever is their
contribution to european infrastructure (I wish I'd be there :-) )
Thus your packet now travel from NSF to D via router R' and eventually
arrives at D
Often R and R' are:
1- connected together
2- both in Europe
3- in charge of funding their infrastructure (no way to spell NSF in Europe
with possible Nordic exception )
1- and 2- plead for a shorter path; making both Michel Fingerhut and NSF
people smile (using US and NSF as transit network is somewhat ridiculous
and not favourably seen by our US collegues (don't read: nastily) )
3- makes R and R' carefull about restricting infrstucture contributed by
group of users to only those groups of users.
It is often the case that S and D belong to different communities funding
different infrastructures without peer agreement between those communities.
Some have already succeeded setting up those peer agreements (read: NORDUnet)
We are working quite hard in RIPE to push more peer agreements
It is more often the case that one (and sometimes both) community does not
contribute for European infrastructure, usually arguing that only
North America is worhtwhile to them (or that a soon to come network will
solve their problems, ...)
Thus that very community uses the "free horse" paid by the other community.
Would you *really* expect that other community (read: it's members through
their contributions) to quietly be such a free horse?, certainly not! thus
that community usually bares access to their infrastructures.
Of course this is not the best solution and "free horse" users have to be
educated and explained that they also must contribute to their own share
of better networking infrstructure in Europe, but here lies many politics
which not always has to do with good sense.
Another example of this situation can be found in central Europe
for motorways and alpine roads within Austria, Germany, Italia,
and Switzerland: for long cars and trucks of a given nation were
allowed to use roads of another nation. The enormous increase in
road traffic and the different pace at which nations have modernised
their motorways (I am pointing at no one precisely) have made
made the financial pressure very unevenly shared among those
countries and resulted in the current crisis where some countries
feel they are overpaying, and are blocking foreign lorries at their
border. (BTW: this has nothing to do with the freedom of circulation)
Other example is the US road system whereby commercial usage of state
roads implies some financial costs for lorries housed in other
states (like being registered, thus the nice trucks or buses with
tens of circulation plates on them). Thus also the Interstate road
system where funded by fede4ral money where it is allowed for any
lory to be involved in business in whatever state!
I hope this add some light to the situation
Yves Devillers
--
Yves Devillers Yves.De...@inria.fr
INRIA Rocquencourt ...uunet!inria!devill
BP 105, 78153 Le Chesnay CEDEX C=Fr/A=/P=/O=inria/S=Devillers/G=Yves
France