[news.admin] Using "local" as a newsgroup name considered harmful

allbery@ncoast.UUCP (Brandon Allbery) (01/01/70)

As quoted from <563@uthub.toronto.edu> by thomson@uthub.toronto.edu (Brian Thomson):
+---------------
| In article <2857@psuvax1.psu.edu> flee@gondor.psu.edu (Felix Lee) writes:
| >Solution:  Make all names unique.  Either make sure they're globally
| >unique, like UUCP, or qualify the names, like ARPA:
| >	/world/na/usa/canada/waterloo/uw.general
| 		  ^^^^^^^^^^
| >Something less clumsy would be better.
| 
| Manifest Destiny, wot?
+---------------

Good point.  Umm, Felix, I hate to say this, but Canada isn't part of the U.S.
(No Reagan, No Ollie North; they've got a good thing going ;-)
-- 
 Brandon S. Allbery, moderator of comp.sources.misc and comp.binaries.ibm.pc
  {{harvard,mit-eddie}!necntc,well!hoptoad,sun!mandrill!hal}!ncoast!allbery
ARPA: necntc!ncoast!allbery@harvard.harvard.edu  Fido: 157/502  MCI: BALLBERY
   <<ncoast Public Access UNIX: +1 216 781 6201 24hrs. 300/1200/2400 baud>>
** Site "cwruecmp" is changing its name to "mandrill".  Please re-address **
*** all mail to ncoast to pass through "mandrill" instead of "cwruecmp". ***

bob@acornrc.UUCP (Bob Weissman) (08/12/87)

I have my sys file set up so that a certain user gets a certain
newsgroup by mail.  I included "local" as a distribution in the sys
line for him.  So far so good.

Today, I noticed (while just idly perusing LIBDIR/log), that this user
was getting other newsgroups (ones he didn't want) in addition to the
one he did want.  Not so good.

It turns out that there's a site called ozdaltx.uucp which uses "local"
not as a distribution name, but as a top-level newsgroup name.  I.e.,
my poor user was getting drivel from "local.general", a newsgroup
local to ozdaltx, because of cross-posting.

Needless to say, I quickly patched my sys file to say "local,!local.all",
but there's a problem here, stemming from the fact that news 2.11 does
not distinguish between distributions and newsgroups in the sys file.

Will the mythical C news distinguish between distributions and top-level
newsgroup names?  Or will we be forever at the mercy of twits who decide
to create local newsgroups called "world.nosepicking" and cross-post to
them?

(Or am I way off base and has every knowledgable news admin already
put "distribution,!distribution.all" in place of just "distribution" in
their sys files?  (Which of course doesn't work for genuine regional
newsgroups like ca.*))

-- 
Bob Weissman
Internet:	bob@acornrc.UUCP
UUCP:		...!{ ames | decwrl | oliveb | apple }!acornrc!bob
Arpanet:	bob%acornrc.UUCP@AMES.ARPA

geoff@utstat.UUCP (08/13/87)

Bob Weissman asks:
> Will the mythical C news distinguish between distributions and top-level
> newsgroup names?

C rnews will indeed support a new sys file field for distributions.  If
this field is missing, rnews will use the contents of the newsgroup
pattern field (`subscription list') as the distribution list.  (I was
mildly surprised that B2.11 didn't include a distribution field, as
Rick is aware of the problem.)

C news isn't mythical, just slow to be released.  It really is running
on utstat and utzoo (and at certain alpha test sites), but Henry and I
aren't getting paid to write news software, so C news is necessarily
getting done in our copious spare time.
-- 
Geoff Collyer	utzoo!utstat!geoff, utstat.toronto.{edu,cdn}!geoff

scott@its63b.UUCP (08/14/87)

In article <434@acornrc.UUCP> bob@acornrc.UUCP (Bob Weissman) writes:
>It turns out that there's a site called ozdaltx.uucp which uses "local"
>not as a distribution name, but as a top-level newsgroup name.  I.e.,
>my poor user was getting drivel from "local.general", a newsgroup
>local to ozdaltx, because of cross-posting.

We have a local.general newsgroup too, and it turns out that the
article in question, namely:

+--
| From: root@ozdaltx.UUCP (root)
| Newsgroups: sci.med,local.general
| Subject: STRESS DIET :-)
| Message-ID: <4157@ozdaltx.UUCP>
+--

gets to our site (since we get sci.med) and then inews inserts it into
both newsgroups. So _our_ local newsgroup gets anything posted to
someone else's local group that's cross-posted to a worldwide group!

I still run 2.10.3 news, which removes unknown newsgroups from the
Newsgroups: line. This feature was removed in 2.11 I think. Perhaps it
should be reinstated for "local.*". The only other thing I can think
of offhand is to stop crossposting between local.* and anything else.

Presumably the same effect takes place in crossposting between
"general" and something else, on sites that have "general" instead of
"local.general".

Scott Larnach
Edinburgh University Computer Service

javoskamp@watmath.waterloo.edu (Jeff Voskamp) (08/16/87)

In article <587@its63b.ed.ac.uk> scott@its63b.ed.ac.uk (Scott Larnach) writes:
>We have a local.general newsgroup too, and it turns out that the
>article in question, namely:
>+--
>| From: root@ozdaltx.UUCP (root)
>| Newsgroups: sci.med,local.general
>| Subject: STRESS DIET :-)
>| Message-ID: <4157@ozdaltx.UUCP>
>+--
>gets to our site (since we get sci.med) and then inews inserts it into
>both newsgroups. So _our_ local newsgroup gets anything posted to
>someone else's local group that's cross-posted to a worldwide group!
>
>I still run 2.10.3 news, which removes unknown newsgroups from the
>Newsgroups: line. This feature was removed in 2.11 I think. Perhaps it
>should be reinstated for "local.*". The only other thing I can think
>of offhand is to stop crossposting between local.* and anything else.

That still won't solve the problem completely (removing local.* groups
that is).  A few months back everyone here was surprised to see an
posting for a gay/lesbian dance show up in the local group uw.general.
The surprise was not that the dance was advertized, but that it was at
the University of Wisconsin.  It seems that the dance was cross posted
to soc.motss and uw.general (i.e. University of Wisconsin general
announcements), wandered across the country in soc.motss and fell into
our uw.general.

One solution would be to have some option for news that will remove a
selected list of news groups, not all unknown ones (a file like
/usr/lib/news/localgroups or the like).  In this way you could remove
all the groups which are local to your site and still pass along news
that is cross-posted to several groups, only some of which you
offically receive. You might have to define which groups get cut out on 
a machine by machine basis, but once it was set up you could forget
about it.

Jeff Voskamp

Opinions are my own. As to software - I don't know, I've never seen the source.
-- 
This should be fine, but I wouldn't put my life on the line over it.
bang path: {allegra,decvax,utzoo,clyde}!watmath!javoskamp 
domain   : javoskamp@watmath.waterloo.{edu,csnet,cdn}
CSNET    : javoskamp%watmath@waterloo.CSNET

piet@cwi.nl (Piet Beertema) (08/17/87)

	>I still run 2.10.3 news, which removes unknown newsgroups from the
	>Newsgroups: line. This feature was removed in 2.11 I think. Perhaps it
	>should be reinstated for "local.*". The only other thing I can think
	>of offhand is to stop crossposting between local.* and anything else.
Local or organizational newsgroups may have almost any name,
not necessarily local.whatever; special treatment of local.*
therefore is no solution. The same problem could occur when
posting to a newsgroup but giving it a distribution "wider"
than that of the newsgroup.
Maybe the news software should refuse any such posting; but
then it would have to know exactly what the distribution
"levels" are.

-- 
	Piet Beertema, CWI, Amsterdam
	(piet@cwi.nl)

root@hobbes.UUCP (08/17/87)

+---- Jeff Voskamp writes in article <14347@watmath.waterloo.edu> ----
| One solution would be to have some option for news that will remove a
| selected list of news groups, not all unknown ones (a file like
| /usr/lib/news/localgroups or the like).  In this way you could remove
| all the groups which are local to your site and still pass along news
| that is cross-posted to several groups, only some of which you
| offically receive. You might have to define which groups get cut out on 
| a machine by machine basis, but once it was set up you could forget
| about it.
+----

Wouldn't there be enough info in the sys file + active file to determine
this?  ie:
	local = in(active) and not in("L" flagged sys file entries)

If I have a local group, (by definition) it is not sent to other machines.
If the message was for internal consumption at a company, the gateway machine
which fed the rest of the net would not have the local groups sent out.

Too farfetched?  Did I forget some simple fact?  flame me by email,
only post constructive, mature replies.

-- 
John Plocher uwvax!geowhiz!uwspan!plocher  plocher%uwspan.UUCP@uwvax.CS.WISC.EDU

dsp@ptsfa.UUCP (David St.Pierre) (08/18/87)

In article <328@sering.cwi.nl> piet@cwi.nl (Piet Beertema) writes:
>
>Local or organizational newsgroups may have almost any name,
>not necessarily local.whatever; special treatment of local.*
>therefore is no solution. The same problem could occur when
>posting to a newsgroup but giving it a distribution "wider"
>than that of the newsgroup.
>Maybe the news software should refuse any such posting; but
>then it would have to know exactly what the distribution
>"levels" are.

Or maybe the news software could be a "tad" smarter about selecting
distributions. It currently seems to select the default distribution
based on the first newsgroup only.

Posting to pb.general and news.misc would give me a distribution of "pb",
our local newsgroups.

Posting to news.misc and pb.general gives me a default distribution of
"world".

If the user posts to multiple newsgroups and the default distribution
for any of them is not equal to "world", let that be the default. This 
basically goes back to regional distributions which is often what people
wanted in that case.  I don't think it's worth trying to make people post
to the most restrictive newsgroup first. People tend to think of a "natural"
category first (soc.motss) and a local one second - or at least some people
seem to.

I don't think the issue should be in refusing a posting, but rather
making the first choice the more restrictive one. Many people use the
default distribution without even looking at it.

-- 
David St. Pierre    415/823-6800  {ihnp4,lll-crg,ames,pyramid}!ptsfa!dsp
He smokes like a fish.

henry@utzoo.UUCP (Henry Spencer) (08/18/87)

> The surprise was not that the dance was advertized, but that it was at
> the University of Wisconsin.  It seems that the dance was cross posted
> to soc.motss and uw.general...

U of Toronto occasionally sees U of Texas news too.  Geoff and I are thinking
about a solution for this for C news; it's not simple.
-- 
Support sustained spaceflight: fight |  Henry Spencer @ U of Toronto Zoology
the soi-disant "Planetary Society"!  | {allegra,ihnp4,decvax,utai}!utzoo!henry

flee@gondor.psu.edu.UUCP (08/19/87)

In article <8430@utzoo.UUCP> henry@utzoo.UUCP (Henry Spencer) writes:
>U of Toronto occasionally sees U of Texas news too.  Geoff and I are thinking
>about a solution for this for C news; it's not simple.

Case:  Someone at Univ. Wisconsin posts to soc.motss and uw.general,
distribution world.  The article turns up in uw.general at Univ. Waterloo.

Case:  I want to post to Penn State and Univ. Waterloo, so I post the
article to psu.general and uw.general, distribution na.  It turns up
in uw.general at Univ. Wisconsin.

Problem:  Given two regions A1 and A2 with the same name A, how do you
tell if the name A refers to A1 or A2?

Answer:  You can't.

Solution:  Make all names unique.  Either make sure they're globally
unique, like UUCP, or qualify the names, like ARPA:
	/world/na/usa/canada/waterloo/uw.general
Something less clumsy would be better.

Solution:  Disallow posting to a region you don't belong in.  Force
Distribution to be the smallest distribution of all the Newsgroups,
so the article doesn't ever leave the region.

(Currently, a newsgroup's distribution is the top-level name, if that
name is found in LIBDIR/distributions; otherwise, it's "world".)

Munging the Newsgroups line to delete local groups won't work, because
the article could conceivably hop out of the region and hop back in
faster than it hops across the region.

Any other solutions?
--
Felix Lee	flee@gondor.psu.edu	{cbosgd,cmcl2}!psuvax1!gondor!flee
Copyright (C) 1987 by Felix Lee.  Reprinted with permission.

nyssa@terminus.UUCP (The Prime Minister) (08/19/87)

In article <8430@utzoo.UUCP> henry@utzoo.UUCP writes:
>U of Toronto occasionally sees U of Texas news too.  Geoff and I are thinking
>about a solution for this for C news; it's not simple.

I'd suspect that you could develop a simple mechanism for the
news software to recognize top level distributions, and pass those
along.  You just remove from the newsgroup line those groups that
aren't top level.

Of course, not knowing the design of C news, that may be harder than
it sounds.

roy@phri.UUCP (Roy Smith) (08/19/87)

In article <8430@utzoo.UUCP> henry@utzoo.UUCP (Henry Spencer) writes:
> U of Toronto occasionally sees U of Texas news too.  Geoff and I are thinking
> about a solution for this for C news; it's not simple.

	Seems to me that "which ut usenet distribution do you mean?" is
exactly the same as the "which zeus uucp site do you mean?" problem that
plagues uucp mail auto-routers.  Both types of name only make sense
relative to a particular site.  Anybody for registered usenet domains?
-- 
Roy Smith, {allegra,cmcl2,philabs}!phri!roy
System Administrator, Public Health Research Institute
455 First Avenue, New York, NY 10016

gore@nucsrl.UUCP (Jacob Gore) (08/19/87)

/ nucsrl:news.admin / flee@gondor.psu.edu (Felix Lee) / 10:53 pm  Aug 18, 1987 /
> [...]
>Solution:  Make all names unique.  Either make sure they're globally
>unique, like UUCP, or qualify the names, like ARPA:
>	/world/na/usa/canada/waterloo/uw.general
>Something less clumsy would be better.
>
>Solution:  Disallow posting to a region you don't belong in.  Force
>Distribution to be the smallest distribution of all the Newsgroups,
>so the article doesn't ever leave the region.
> [...]
>Any other solutions?
>--
>Felix Lee	flee@gondor.psu.edu	{cbosgd,cmcl2}!psuvax1!gondor!flee
>Copyright (C) 1987 by Felix Lee.  Reprinted with permission.

Sure.  Just don't cross-post something that goes to a global group to any of
the local groups.

Jacob Gore				gore@EECS.NWU.Edu
Northwestern Univ., EECS Dept.		{gargoyle,ihnp4,chinet}!nucsrl!gore

thomson@uthub.UUCP (08/20/87)

In article <2857@psuvax1.psu.edu> flee@gondor.psu.edu (Felix Lee) writes:
>Solution:  Make all names unique.  Either make sure they're globally
>unique, like UUCP, or qualify the names, like ARPA:
>	/world/na/usa/canada/waterloo/uw.general
		  ^^^^^^^^^^
>Something less clumsy would be better.

Manifest Destiny, wot?
-- 
		    Brian Thomson,	    CSRI Univ. of Toronto
		    utcsri!uthub!thomson, thomson@hub.toronto.edu

david@ms.uky.edu (David Herron -- Resident E-mail Hack) (08/21/87)

In article <8430@utzoo.UUCP> henry@utzoo.UUCP (Henry Spencer) writes:
>> The surprise was not that the dance was advertized, but that it was at
>> the University of Wisconsin.  It seems that the dance was cross posted
>> to soc.motss and uw.general...
>U of Toronto occasionally sees U of Texas news too.  Geoff and I are thinking
>about a solution for this for C news; it's not simple.

Yes Henry ... please do come up with a solution.  We've recently gotten
a small spate of messages from a person in England crossposting things
between something called "uk.general" and other newsgroups.  We've got
a "uk.general" and some other uk.whatevers where uk stands for U of Kentucky.
They've got a uk.general (I suppose) where uk stands for United Kingdom.

The problem is general enough and we're seeing enough postings about it,
that obviously a solution needs to be found.

With the current set of headers and current sys file and active file, I
can't see any solution other than removing newsgroups which aren't known.
(and moving the article into junk if it there aren't any newsgroups
after that).  But that has repercussions where you have an article
originating at one site cross-posted to newsgroups known to A and C
but some are not known by B.  If the article happens to pass through 
B to C before it does A to C, then the one on C will not appear in
all the newsgroups it's supposed to appear in.  (unless you keep the
newsgroup name in the Newsgroups: line, BUT we have our current problem
if we do that ...)

sigh
-- 
----- David Herron,  Local E-Mail Hack,  david@ms.uky.edu, david@ms.uky.csnet
-----                            {uunet,cbosgd}!ukma!david, david@UKMA.BITNET
----- 
----- Je parle francais comme une vache espagnole.

javoskamp@watmath.waterloo.edu (Jeff Voskamp) (08/22/87)

In article <2857@psuvax1.psu.edu> flee@gondor.psu.edu (Felix Lee) writes:
>Case:  Someone at Univ. Wisconsin posts to soc.motss and uw.general,
>distribution world.  The article turns up in uw.general at Univ. Waterloo.
>
>Case:  I want to post to Penn State and Univ. Waterloo, so I post the
>article to psu.general and uw.general, distribution na.  It turns up
>in uw.general at Univ. Wisconsin.
Actually I don't think it will.  From what I know news is only shipped out
if at least one of the newsgroups is in the L.sys file.  News posted to 
uw.general and soc.motss will get out because of the soc.motss and for that
reason only.  News posted to uw.general only will not leave the university
since we don't ship uw.general to anyone.  Since uw.* and psu.* aren't in 
the L.sys file here the article won't go anywhere.

>Problem:  Given two regions A1 and A2 with the same name A, how do you
>tell if the name A refers to A1 or A2?
>
>Answer:  You can't.
Too true.
>
>Solution:  Make all names unique.  Either make sure they're globally
>...
>Munging the Newsgroups line to delete local groups won't work, because
>the article could conceivably hop out of the region and hop back in
>faster than it hops across the region.
A question: would the local group hop out of the region and back in if the
article weren't cross posted?  If the answer is no then I'd say, chop out 
the local group name.  However this will cause some problems in the following
case:
given 3 sites a, b, and c where a, and b are in some region p and c is not.
Connections are as follows:
the pairs (a,b) exchange newsgroups x and y (x is local to region p).
(c,b), and (c,a) exchange newsgroup y. 
Problem: (c,a), and (c,b) exchange news every hour. (a,b) exchange news
one per day.  
Someone crossposts an article to x and y from a. As the news goes to c x is 
dropped from the newsgroups line.  Several hours later the news is in group y
at all sites, but is only in group x at site a.
Oh well, nothing's perfect.

How about a distribution per newsgroup?  Each newsgroup has a default distri-
bution, and as it crosses a boundary, the appropriate group(s) gets dropped.

Jeff Voskamp

The opinions expressed are correct.  However, they may be based on incorrect
assumptions.  If the assumptions are wrong, please let me know and I won't
use them in the future.
-- 
This should be fine, but I wouldn't put my life on the line over it.
bang path: {allegra,decvax,utzoo,clyde}!watmath!javoskamp 
domain   : javoskamp@watmath.waterloo.{edu,csnet,cdn}
CSNET    : javoskamp%watmath@waterloo.CSNET

henry@utzoo.UUCP (Henry Spencer) (08/25/87)

Making the group names unique would be the preferred solution in some
ways, but this is hard to retrofit onto the entire universe.  It also
would be a considerable pain to have to type world.na.can.ont.tor.ut.vlsi
to get the UofT VLSI newsgroup.  (This is arguably a user-interface
issue, but it's one that can't be ignored when the current user-interface
software doesn't do anything intelligent about it.)

Unfortunately, it is difficult to devise another solution that really
works without unfortunate side effects.  Stripping inappropriate groups
out of the Newsgroups line causes trouble when news can loop out of a
"local newsgroup area" and back in; this applies equally to stripping
when the article leaves the area and to stripping when it enters from
outside.  (In addition, Geoff and I have strong religious prejudices
against header alteration of any kind, and would prefer not to add more
of it on top of the bits that rnews already has to do.)  Leaving the
inappropriate groups in the header but not posting the article to them
is fine for the sites at the border of the local area, who know that the
article came in from outside, but doesn't work for sites farther in.
And again it does the wrong thing when news loops via outside sites.
Restricting the distribution of cross-posted stuff to the smallest common
region would work, sort of, once the restriction code was widespread, but
makes life awkward when one really *wants* to post something to both a
local group and a global one (for example, a site-is-defunct notice to
both ont.uucp and news.config).  This may be the least of the evils,
though.

Any other ideas?
-- 
"There's a lot more to do in space   |  Henry Spencer @ U of Toronto Zoology
than sending people to Mars." --Bova | {allegra,ihnp4,decvax,utai}!utzoo!henry

dave@lsuc.UUCP (08/26/87)

Henry Spencer (utzoo!henry) writes:
>Unfortunately, it is difficult to devise another solution that really
>works without unfortunate side effects.  Stripping inappropriate groups
>out of the Newsgroups line causes trouble when news can loop out of a
>"local newsgroup area" and back in...  Leaving the
>inappropriate groups in the header but not posting the article to them
>is fine for the sites at the border of the local area, who know that the
>article came in from outside, but doesn't work for sites farther in...
>Any other ideas?

Yeah. What about looking at the domain of the poster in the From:
line when you receive the article, and making a decision based
on that as to whether to post to the local group a cross-posted article
which includes a locallish-distribution group? (For sanity protection,
don't invoke this test at all unless the article is cross-posted to
both a wide and a narrow distribution.)

To keep this information current, have a file with "known"
problems. For example, U of Toronto sites would have, in the "zap this"
file,
ut	.texas.edu
meaning, "if something comes in cross-posted to ut.anything,
and it was posted from anyone@anything.texas.edu, then don't post
it to ut.anything because they meant a different ut."
Similarly, the people at Kentucky would have
uk	.uk
while those in Britain would have
uk	.uky.edu

The file would be updated by the sysadmins every time
a new duplication problem surfaced.

The only problem I see is that if joe@texas.edu posts to
ut.general,comp.misc, and fred@toronto.edu reads the article
in comp.misc, the header will still mention ut.general (thus
preserving the article in case it goes back into the Texas
distribution, in theory), but fred will not be able to find
it there if he goes looking. Not a serious problem, in my view.

David Sherman
The Law Society of Upper Canada
Toronto
-- 
{ uunet!mnetor  pyramid!utai  decvax!utcsri  ihnp4!utzoo } !lsuc!dave

david@ms.uky.edu (David Herron -- Resident E-mail Hack) (08/26/87)

In article <2007@lsuc.UUCP> dave@lsuc.UUCP (David Sherman) writes:
>Henry Spencer (utzoo!henry) writes:
>> ... Unfortunately, it is difficult to devise another solution that really
>>works without unfortunate side effects.   ...
>Yeah. What about looking at the domain of the poster in the From:
>line when you receive the article, and making a decision based
>on that as to whether to post to the local group a cross-posted article
>which includes a locallish-distribution group? (For sanity protection,
>don't invoke this test at all unless the article is cross-posted to
>both a wide and a narrow distribution.)
>
>To keep this information current, have a file with "known"
>problems. For example, U of Toronto sites would have, in the "zap this"
>file,
>ut	.texas.edu
> ... Similarly, the people at Kentucky would have
>uk	.uk
>while those in Britain would have
>uk	.uky.edu
>
>The file would be updated by the sysadmins every time
>a new duplication problem surfaced.

I don't really like this solution as it is the same sort of thing
as having to know the paths around the network ... the network is
getting too big in order for us to know what all is out there.
(BTW, it's "utexas.edu" not "texas.edu" :-) ...).

In addition, not all of the distributions are attached to particular
domains ...

For instance, in KY we have a "ky" distribution.  This distribution
crosses a number of domains.  There is also the "alt" distribution
which crosses a number of domains.

If the maker of "ky-jelly" were to make an internal "ky" distribution
then we might start seeing crosspostings here.  :-)





[I'm just being devils advocate at the moment and pointing at some
 problems ... I also realize that we've only ever had one bout of
 accidental cross-postings into our internal newsgroups in the 2 yrs
 or so we've had internal newsgroups.  I could certainly live with
 making one entry into a file to cover further cross-postings.

 On the other hand, I wonder if people in the United Kingdom will
 see this posting as I've cross-posted it into uk.news.  The "uk.news"
 group in the UK is probably much busier than our local newsgroups.
 
 My experience with our local newsgroups is probably the exception
 rather than the rule ... For awhile we were getting leakage from
 internal newsgroups at GaTech and was seeing quite a bit of
 traffic in those newsgroups ... a whole lot more than our internal
 newsgroups get.]




>The only problem I see is that if joe@texas.edu posts to
>ut.general,comp.misc, and fred@toronto.edu reads the article
>in comp.misc, the header will still mention ut.general (thus
>preserving the article in case it goes back into the Texas
>distribution, in theory), but fred will not be able to find
>it there if he goes looking. Not a serious problem, in my view.

Neither in my view either.
-- 
----- David Herron,  Local E-Mail Hack,  david@ms.uky.edu, david@ms.uky.csnet
-----                            {uunet,cbosgd}!ukma!david, david@UKMA.BITNET
----- 
----- Je parle francais comme une vache espagnole.

wedgingt@udenva.UUCP (Will Edgington) (08/29/87)

In article <2007@lsuc.UUCP> dave@lsuc.UUCP (David Sherman) writes:
>Yeah. What about looking at the domain of the poster in the From:
>line when you receive the article, and making a decision based
>on that as to whether to post to the local group a cross-posted article
>which includes a locallish-distribution group? [...]
>
>[...] For example, U of Toronto sites would have, in the "zap this"
>file,
>ut	.texas.edu
>meaning, "if something comes in cross-posted to ut.anything,
>and it was posted from anyone@anything.texas.edu, then don't post
>it to ut.anything because they meant a different ut." [...]

How about reversing the meaning of the domain ?  I.e., a line like :

ut	.texas.edu

means *keep* 'ut.' newsgroups if the poster's domain ends with '.texas.edu' ?

>The file would be updated by the sysadmins every time
>a new duplication problem surfaced.

My variation obviates the need for this, and, furthermore, you don't need
to know of a problem to have the file ready to squash it.  You only need
to change it when you add a new distribution.  Better yet, you can now
expand the file to handle cases like :

co	.du.edu,.ucar.edu,.isc.com,.nbi.com,.alphacdc.com

where the 'co.' distribution is the same one for all of the given (mail)
domains (yes, .du.edu and .ucar.edu *aren't* in the UUCP maps, and these
*are* most of the domains I know of in Colorado).

>The only problem I see is that if joe@texas.edu posts to
>ut.general,comp.misc, and fred@toronto.edu reads the article
>in comp.misc, the header will still mention ut.general (thus
>preserving the article in case it goes back into the Texas
>distribution, in theory), but fred will not be able to find
>it there if he goes looking. Not a serious problem, in my view.

Nor in many others, including mine, from what I've heard ...
Part of the "Frequently Asked Questions" document could even
say a bit about this kind of thing.

Is this too late for C News ?  If so, is it small enough changes
to be sent out as an "enhancement patch" later on ?
-----
Will Edgington, Computing and Information Resources, University of Denver
System Administrator for udenva (== dueos), dutyche, duorion, dunike, ...
{{hplabs,seismo}!hao,ucbvax!nbires,boulder,cires,ssds}!udenva!wedgingt,
WEDGINGT@DUCAIR (BITNET), wedgingt@ccndu (CSN/CCN), ...
COMING SOON: wedgingt@du.edu (all nets)|"No two addresses are the same  (Ail

henry@utzoo.UUCP (Henry Spencer) (08/31/87)

> Yeah. What about looking at the domain of the poster in the From:
> line when you receive the article, and making a decision based
> on that as to whether to post to the local group a cross-posted article
> which includes a locallish-distribution group? ...

Geoff and I kicked this idea around a bit, but couldn't find a workable
variation.  The problem is that the From: line format varies and does
not dependably contain a particular domain.  For example, this article
is from "henry@utzoo.UUCP" (unless something along the way has mucked
up the From: line), and rules applying to "toronto.edu" won't catch it.
Worse, even if we assume everyone updates to using proper domains (dream
on...), the correspondence between domains and local news regions is at
best coincidental, which shoots down fancier versions of the scheme.
For example, there is no domain or small set of domains that corresponds
to the "ont" region.
-- 
"There's a lot more to do in space   |  Henry Spencer @ U of Toronto Zoology
than sending people to Mars." --Bova | {allegra,ihnp4,decvax,utai}!utzoo!henry

chris@gargoyle.UChicago.EDU (Chris Johnston) (09/05/87)

There is nothing you can do as long as you restrict the name space to
2 or 3 characters.  If you use names like utexas.wanted or
uwisc.general or bayarea.eats you won't have a problem.

cj