steve@mrc-apu.cam.ac.uk (Steve Platt) (06/13/89)
|Shouldn't this order be changed?
Maybe, although it HAS matched a higher level name first;
however, if I had a host/sub-department called "Acorn"
I'd think twice before registering it in that name under
the NRS.
As things in the NRS often seem like words (which might also be
proper names) and since computervision isnt a word (although
it is a proper name as you found) couldn't you elect to register
as uk.ac.bristol.computer-vision (uk.ac.bris.cv)?
Mind you, the prize must stay with the Janet Central office
for the UK.AC.CO masterpiece; Internet worm, eat-your-heart-out!HopkinsRJ@bristol.ac.uk (R.J.Hopkins) (06/14/89)
I'm in the process of adding a new mail host to the NRS. It will be known as uk.ac.bristol.computervision (uk.ac.bris.cvn). I've found that with sendmail2.1, an address of the form "fred@computervision" resolves to fred@uk.co.computervision because of the order of searches in Ruleset 11 (uk, uk.ac, uk.co, uk.ac.bristol in my case). Shouldn't this order be changed? Richard Hopkins
andrew@stl.stc.co.uk (Andrew Macpherson) (06/14/89)
| >Shouldn't this order be changed? | | Maybe, although it HAS matched a higher level name first; Precicely, this is a mistake. To be consistent with name-server operation, and the "domain names, concepts and facilities" RFC, matching should be from local outward. If you look at the resolver code (or set debug in nslookup) you will see it try successively shorter parts pf your local domain before submitting the "unadorned" string for a match. The idea is to allow you to use a "short-form" to describe relatively local entities, and only to require you to specify *from the point of divergance with your local domain* for more distant. Reversing this would be crazy. If I wanted to call a machine "uk.co.stc.stl.mod", [ ie "rocker"'s friend] I expect to be able to refer to it as "mod", and not wind up with mail being sent to RSRE. Simmilarly, if STC Telecoms had that machine, I would expect to specify "tcom.mod"; this is where the UK sendmail has been "too nice" to the users in the past... now that I am using the name-server code I can no longer get away with the short-form in the non-local domains. (All addresses reversed to conform with NRS strangenesses) | however, if I had a host/sub-department called "Acorn" | I'd think twice before registering it in that name under | the NRS. | I'd be very tempted to register one of our hosts, just to make sure everyone's software is resiliant "uk.co.stc.icl.rbs.janet" :-) | couldn't you elect to register | as uk.ac.bristol.computer-vision (uk.ac.bris.cv)? | [long and boring diatribe about excessively long and un-typeable names suppressed] -- -- Andrew Macpherson PSI%234237100122::andrew andrew@stl.stc.co.uk - or - ...!mcvax!ukc!stl!andrew "It is always a great mistake to treat the individual on the chance that he may become a crowd" -- Mr Justice Codd: (A.P.Herbert)
jac@doc.ic.ac.uk (Jim Crammond) (06/14/89)
The existing ordering allows you to say user@computervision and see it expand to user@computervision.co.uk and not user@computervision.co.ac.uk ^^ As Steve Platt points out, we still have this problem with co.ac.uk being registered. If you want computervision to map to something local then the obvious answer is to remove the *ambiguity* by taking it out of your table of co.uk sites; or taking out the co.uk table altogether. -Jim.
jim@cs.strath.ac.uk (Jim Reid) (06/19/89)
|> We in fact do something more - a makefile generates new NRS derived |> input decks by editing out names we know to be a problem, like 'uk.ac.co', |> 'uk.ac.mrc' etc (we have a uk.ac.gla.mrc) This is an unsightly (and strictly speaking irrelevant) kludge. The whole basis of the NRS was to prevent ambiguous domain names. The rules say that a domain name cannot be used if it is already in use as a higher-level name within the domain hierarchy. If that rule was applied (oh that it were...), names like 'uk.ac.gla.mrc' and 'uk.ac.co' could not be used since they caused ambiguities. [Does mail for 'co.foo' go to 'uk.co.foo' or 'uk.ac.co.foo'?] They wouldn't then be permitted to appear in the NRS database and there would be no need for people to kludge their mail tables. Why does the NRS let people get away with registering illegal addresses? Of course, everybody knows that real-world mail systems need quick and dirty hacks like that since reality tends to be different from what the standards say..... Jim
taylor@cs.glasgow.ac.uk (Jem Taylor) (07/22/89)
|Think we're talking at cross purposes. Below is a script listing showing |the creation of 2 configs - test1 and test2. They are identical except |for the order of the domain declarations in the config file. You'll |see that there are indeed differences between the resultant sendmail |configuration files but these diffences do not affect the order of |domain resolution. In both, "computervision" still resolves to |computervision.co.uk I see your point. I suggest you do as the manual tells you, and use the sequence (assume your config is called $config ) $ cd $SENDMAIL/$config && make check $config install This will abort if any potential conflicts are detected. You can then delete the entry you dislike. We in fact do something more - a makefile generates new NRS derived input decks by editing out names we know to be a problem, like 'uk.ac.co', 'uk.ac.mrc' etc (we have a uk.ac.gla.mrc) I enclose the relevant script, which Jim thought too obvious to put into the package:- mk.sendmail.db :::::::::::::: #!/bin/sh # # Bring sendmail domain databases up-to-date # /bin/make -f - $@ <<'EOF' NRS=/usr/local/src/c-nrs NRSFILES=$(NRS)/ac.dom $(NRS)/top.dom $(NRS)/co.dom $(NRS)/gla.dom DERFIL=/usr/spool/nrs/DERFIL2 SENDMAIL=/usr/local/lib/Sendmail EXTERNAL= eng eng-slave psy psy-slave geology geol-slave science science-slave\ art art-slave all: $(DERFIL) hawaii crete slave $(EXTERNAL) getfiles hawaii crete: $(NRSFILES) cd $(SENDMAIL)/$@ && make check $@ install external: $(EXTERNAL) $(EXTERNAL): $(NRSFILES) cd $(SENDMAIL)/$@ && make check $@ slave: cd $(SENDMAIL)/$@ && make $@ check $@ && ./install-$@ $(DERFIL): $(DERFIL).Z -uncompress -f $(DERFIL) $(DERFIL).Z: getfiles: cd $(NRS) && ./$@ $(NRSFILES): $(DERFIL) cd $(NRS) && make sendmail for i in $(NRS)/*.dom $(NRS)/*.chn ; \ do sort -u $$i | sed -e '/^co\./s/^/#/' \ -e '/uk\.ac\.gla\.cs/s/^/#/' \ -e '/uk\.ac\.glasgow\.cs/s/^/#/' \ -e '/uk\.co\.stc/s/^/#/' \ -e '/^mrc /s/^/#/' \ -e '/^ac\./s/^/#/' >$$i-t \ && mv -f $$i-t $$i \ && chmod a+r $$i ; \ done EOF -- Arpa:taylor@cs.glasgow.ac.uk \ J.A.Taylor, Computing Science, Janet: taylor@uk.ac.glasgow.cs \ University of Glasgow, Uucp: mcvax!cs.glasgow.ac.uk!taylor \ GB-GLASGOW G12 8QQ
taylor@cs.glasgow.ac.uk (Jem Taylor) (07/22/89)
I'm in the process of adding a new mail host to the NRS. It will be known as uk.ac.bristol.computervision (uk.ac.bris.cvn). I've found that with sendmail2.1, an address of the form "fred@computervision" resolves to fred@uk.co.computervision because of the order of searches in Ruleset 11 (uk, uk.ac, uk.co, uk.ac.bristol in my case). Shouldn't this order be changed? Richard Hopkins The order of domain resolution searches is directly dictated by the order of domain declarations in your config.site file. -Jem.
HopkinsRJ@bristol.ac.uk (R.J.Hopkins) (07/23/89)
| The order of domain resolution searches is directly dictated by the order | of domain declarations in your config.site file. | | -Jem. | Think we're talking at cross purposes. Below is a script listing showing the creation of 2 configs - test1 and test2. They are identical except for the order of the domain declarations in the config file. You'll see that there are indeed differences between the resultant sendmail configuration files but these diffences do not affect the order of domain resolution. In both, "computervision" still resolves to computervision.co.uk Here's the script listing... Script started on Fri Jun 16 15:41:24 1989 1## Config config.test1 creating tables/test1.mk ... Making directory test1 ... creating configuration files in test1 ... 2## Config config.test2 creating tables/test2.mk ... Making directory test2 ... creating configuration files in test2 ... 3## cd test1 4## make cd ../tables; make -f test1.mk test1 ../Dombuild -o test1-dom.m4 bristol.local.dom bristol.bris.dom ac.dom uk.dom co.dom -top top.dom ../Chnbuild -o test1-chn.m4 -local qgb.local.chn -ether ether.chn -janet janet.chn -janet catchall.chn -janet bristol.janet.chn -top top.dom make test1.cf cp ../tables/test1-dom.m4 . cp ../tables/test1-chn.m4 . mv: test1.cf: Cannot access: No such file or directory *** Exit 1 (ignored) m4 test1.mc > test1.cf 5## cd ../test2 6## make cd ../tables; make -f test2.mk test2 ../Dombuild -o test2-dom.m4 uk.dom ac.dom bristol.bris.dom co.dom bristol.local.dom -top top.dom ../Chnbuild -o test2-chn.m4 -local qgb.local.chn -ether ether.chn -janet janet.chn -janet catchall.chn -janet bristol.janet.chn -top top.dom make test2.cf cp ../tables/test2-dom.m4 . cp ../tables/test2-chn.m4 . mv: test2.cf: Cannot access: No such file or directory *** Exit 1 (ignored) m4 test2.mc > test2.cf 7## diff test2.cf ../test1/test1.cf 17c17 < ### test2 configuration file --- | ### test1 configuration file 331c331 < ##### @(#)test2-dom.m4 UK-2.1 sendmail configuration 18/11/88 --- | ##### @(#)test1-dom.m4 UK-2.1 sendmail configuration 18/11/88 340c340 < CAalv alvey forest-rd mod ac co --- | CAac co alv alvey forest-rd mod 348c348 < CBabdn aber aberdeen aberystwyth afrc aston bangor bath bbk bham --- | CBbris abdn aber aberdeen aberystwyth afrc aston bangor bath bbk bham 376c376 < CBumrcc uwcm warwick warwk westfield wolverhampton wolves york bris --- | CBumrcc uwcm warwick warwk westfield wolverhampton wolves york 396,397c396,397 < CEbsa cms computervision compsci csa esa fagg gsa lib msa np1a pva qgb < CEqsa qvc rsa lloyd knott testpc domain apl tango --- | CElloyd knott testpc domain apl tango bsa cms computervision compsci csa | CEesa fagg gsa lib msa np1a pva qgb qsa qvc rsa 404d403 < R$*med-image.compsci $1med-image.compsci.bristol.ac.uk 408a408 | R$*med-image.compsci $1med-image.compsci.bristol.ac.uk 452c452 < ##### @(#)test2-chn.m4 UK-2.1 sendmail configuration 18/11/88 --- | ##### @(#)test1-chn.m4 UK-2.1 sendmail configuration 18/11/88 8## script done on Fri Jun 16 15:43:31 1989