isaac@mungunni.OZ (Isaac Balbin) (09/05/85)
I wonder if there is sufficient support for organising a standard user
like the existing "postmaster" on some Unix Systems, which will enable people
to find out the login id of others. I have often sighed when asked "how can I
find out so ans so's login id when he is on an (for example) IBM machine on
the BITNET". This is especially hard when you are not on such a network
directly. Is there already a special user for VMS, MVS, VM systems?
Who are they? Does *everyone* use them? If not, and we all agree that such
a concept is desirable (given that many sites don't even return invalid mail)
How can we effectively enforce such a standard?
On another point, how often have you wished there was a way to find out that
your mail actually got to where you wanted it too - this is especially
desirable when trying out a new path initially. Can such a "notification"
option exist so that one realises that the only reason no reply has been
received is that the receipient is away or uninterested or trying to avoid
answering you!
PS Many smart sendmails wont let you put in specific loops - whatsmore this
isn't always possible anyway.
Any thoughts? ( I apologise if these have been discussed before - I have
not seen it.)
Isaac Balbin
{seismo,mcvax,ukc,ubc-vision}!munnari!isaac
munnari!isaac@seismo.ARPA
isaac@munnari.oz.CSNET
"If I knew him, I would be him"fair@ucbvax.ARPA (Erik E. Fair) (09/07/85)
This idea has been in the netnews documentation for a long time. Most properly set up USENET sites have an address `usenet' which is supposedly read by the person who administrates netnews (I know that the sites that I have had anything to do with are set up this way, and most of the sites that I have had occasion to contact at one time or another in the last three years are similarly equipped). `Postmaster' as a valid mail address is required by the ARPA Internet Mail Standard, RFC822, so all ARPA Internet sites will have a `postmaster', regardless of whether they have netnews or not. I am also told by various reliable sources that much of the BITNET has standardized on `POSTMAST' as a contact ID. Unfortunately for this idea, all of the versions of UNIX, other than the ones that come from Berkeley (2 & 4 BSD), have the most bletcherous mailers that it would ever be your misfortune to encounter, which, without exception, do not support the concept of an `alias' (e.g. ucbvax!usenet, ucbvax!erik, and a few other things, can all point to ucbvax!fair, without existing as entries in /etc/passwd). So until AT&T and its imitators discover what *real* electronic mailing systems are like (and what they do), we're stuck with the world as it is... Erik E. Fair ucbvax!fair fair@ucbarpa.BERKELEY.EDU
guy@sun.uucp (Guy Harris) (09/08/85)
> `Postmaster' as a valid mail address is required by the ARPA Internet Mail > Standard, RFC822, so all ARPA Internet sites will have a `postmaster', > regardless of whether they have netnews or not. 4.xBSD sites (x >= 2) also have "postmaster" as an alias in their /usr/lib/aliases file, whether they're on the internet or not, unless they've deleted it (which they shouldn't do). Sites running a UNIX which comes with "sendmail" may also have "postmaster" in their /usr/lib/aliases file. > Unfortunately for this idea, all of the versions of UNIX, other than > the ones that come from Berkeley (2 & 4 BSD), have the most bletcherous > mailers that it would ever be your misfortune to encounter, which, > without exception, do not support the concept of an `alias' (e.g. > ucbvax!usenet, ucbvax!erik, and a few other things, can all point to > ucbvax!fair, without existing as entries in /etc/passwd). To clarify a little: "mailer" here doesn't refer to the program you use to read and send mail messages (System V has a better mailer called "mailx"; the reason it's better is that it's a hacked-up version of the Berkeley "Mail", although they only mention this fact in comments in the source code - no credit where credit is due...), but to the program which does mail delivery and routing. "Sendmail" and the earlier "delivermail" have an aliasing capability like this. Without getting into the endless debate as to whether "sendmail" is the answer to a maiden's prayer or the horror of the century, having the ability to do such aliasing is extremely useful once your Email system gets above a certain level of complexity. One place where it's *very* useful is if you have a large network of machines and one machine which acts as a gateway to a large mail network, like UUCP or the Internet. You can have everybody mail to, say, "user@sun" and use the alias mechanism to route mail to the user's own machine. This hides the details of your internal network from the rest of the world, who probably doesn't want to know those details. Using "sendmail", you can also rewrite the "From" addresses of outgoing mail so that it looks like it comes from "user@sun" or whatever, thus completely hiding your internal machines from the outside world. (Yes, I know Sun doesn't do that.) Guy Harris
fair@ucbvax.ARPA (Erik E. Fair) (09/08/85)
I did once write a sendmail.cf at DUAL to hide the internal network, and had aliases for each user on the network on the main machine. Once I finished, I was freed from any constraints or worries about the names of the machines on the internal network. Also, since these aliases existed on all the machines, all I had to remember to mail a note to someone was their username. I didn't have to remember which machine they happened to be reading their mail on this week... The big difference between DUAL and SUN is that DUAL had only 6 machines on the internal ethernet, which made the administrative problem tractable by manual procedures. SUN has hundreds of machines on their multiple ethernets, and thus the mechanics of maintaining such a table of aliases becomes a lot messier. It is at that scale that some type of name service becomes important to implement. Erik E. Fair ucbvax!fair fair@ucbarpa.BERKELEY.EDU
guy@sun.uucp (Guy Harris) (09/09/85)
> The big difference between DUAL and SUN is that DUAL had only 6 > machines on the internal ethernet, which made the administrative > problem tractable by manual procedures. SUN has hundreds of machines on > their multiple ethernets, and thus the mechanics of maintaining such a > table of aliases becomes a lot messier. It is at that scale that some > type of name service becomes important to implement. Actually, the part I was referring to when I said "Yes, I know Sun doesn't do that" is hiding the name of internal machines on outgoing mail. We *do* have a huge alias list that forwards all "user@sun" mail to the appropriate machine. That was done using the /usr/lib/aliases file, not with a name server. Guy Harris
root@bu-cs.UUCP (Barry Shein) (09/09/85)
Re: sendmail and SYSV The lit I recently got on AT&T's UNIX SYSV for VM/370 (really Amdahl's port I believe) lists sendmail among the utilities (I would cheer here but I have this nagging love/hate relationship with sendmail, tho it does do the job after some work.) Maybe, just maybe this is a harbinger of things to come. ATTIS? It would certainly solve a lot of problems for us, esp on the 3081. -Barry Shein, Boston University
gnu@l5.uucp (John Gilmore) (09/09/85)
In article <426@mungunni.OZ>, isaac@mungunni.OZ (Isaac Balbin) writes: > I wonder if there is sufficient support for organising a standard user > like the existing "postmaster" on some Unix Systems, which will enable people > to find out the login id of others. For VM/370 systems [BITNET], the equivalent of "root" is "operator". You could try mailing to there. If each site that connects a new site would ensure that "postmaster" and "usenet" aliases exist on the new site, this would help. (One is for trouble with mail, the other for trouble with netnews.) It isn't in general possible to answer questions like "how to I send mail to Joe Blow at Harvard" without calling him up or sending him a postcard, or finding someone already in electronic touch with him. > Can ... a "notification" > option exist so that one realises that the only reason no reply has been > received is that the receipient is away or uninterested or trying to avoid > answering you! This exists for sites running sendmail or other Arpa-compatible mailers. Add a header line "Return-receipt-to: XXX" where XXX is your address, *relative to the receiving site*. In other words, it's the address your friend would have to send to, to get the message to you.
jer@peora.UUCP (J. Eric Roskos) (09/09/85)
> I wonder if there is sufficient support for organising a standard user > like the existing "postmaster" on some Unix Systems, which will enable people > to find out the login id of others. > ... > Any thoughts? We currently use the name "ns" for that purpose. "postmaster" goes to a human user. "ns" goes to a program that reads the message (it assumes that the message consists simply of the name of the person who the user ID is being requested for) and composes a reply consisting of the output from the "finger" command given the message line as an argument with the "short" switch to prevent other information from being supplied (project, etc). It would be better to have a program like the CSnet nameserver to which you can choose to post or not post your personal information; our program currently restricts the number of queries a person can make, and logs all its activity, to detect abuse. PS - "ns" = "Name Server". It is not particularly mnemonic; my main point was that a separate name from "postmaster" should be used. -- Shyy-Anzr: J. Eric Roskos UUCP: ..!{decvax,ucbvax,ihnp4}!vax135!petsd!peora!jer US Mail: MS 795; Perkin-Elmer SDC; 2486 Sand Lake Road, Orlando, FL 32809-7642
isaac@mungunni.OZ (Isaac Balbin) (09/10/85)
In article <82@l5.uucp> gnu@l5.uucp (John Gilmore) writes: ! In article <426@mungunni.OZ>, isaac@mungunni.OZ (Isaac Balbin) writes: ! > I wonder if there is sufficient support for organising a standard user ! > like the existing "postmaster" on some Unix Systems, which will enable people ! > to find out the login id of others. ! For VM/370 systems [BITNET], the equivalent of "root" is "operator". You ! could try mailing to there. ! Thanks for your advice, but it only serves to show how disorganised things are. A previous news item <10298@ucbvax.ARPA> from Erik Fair quotes reliable sources as saying they have standardised to POSTMAST! I don't really care *how* (viz sendmail) one achieves the aliasing - the point is that there is no uniform way (in practice) as yet to address the "postmaster" on any machine. There should be; and everyone should conform. Can/Will it happen? Will everyone keep RFC822 religiously? ! > Can ... a "notification" ! > option exist so that one realises that the only reason no reply has been ! > received is that the receipient is away or uninterested or trying to avoid ! > answering you! ! This exists for sites running sendmail or other Arpa-compatible mailers. ! Add a header line "Return-receipt-to: XXX" where XXX is your address, ! *relative to the receiving site*. In other words, it's the address your ! friend would have to send to, to get the message to you. Thats nice to know. Is there any chance for this to be standardised too? Isaac Balbin munnari!isaac@seismo.ARPA
chris@umcp-cs.UUCP (Chris Torek) (09/10/85)
In article <428@mungunni.OZ> isaac@mungunni.OZ (Isaac Balbin) writes: > In article <82@l5.uucp> gnu@l5.uucp (John Gilmore) writes: > ! For VM/370 systems [BITNET], the equivalent of "root" is "operator". > ! You could try mailing to there. > ! > Thanks for your advice, but it only serves to show how disorganised > things are. A previous news item <10298@ucbvax.ARPA> from Erik Fair (Look, Erik, he spelled it right! :-) ) > quotes reliable sources as saying they have standardised to POSTMAST! I do not know what standards are in use on BITNET; however, the two articles Isaac Balbin quotes make orthogonal points. All ARPA Unix sites will have both a ``root'' and a ``postmaster''. If you are trying to reach someone on such a machine, either address will suffice. (Postmaster is, of course, better for mail queries.) It is not unreasonable that BITNET would have both ``operator'' and ``postmast''. > I don't really care *how* (viz sendmail) one achieves the aliasing > - the point is that there is no uniform way (in practice) as yet > to address the "postmaster" on any machine. There should be; and > everyone should conform. Can/Will it happen? Will everyone keep > RFC822 religiously? The nature of the UUCP network implies that the answer is ``no''. I do believe, however, that the majority of sites can and will adopt any reasonable standard, as long as the standard is well known. (Definitions of ``reasonable'' may vary, but I think ``postmaster'' will fit into most.) -- In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 4251) UUCP: seismo!umcp-cs!chris CSNet: chris@umcp-cs ARPA: chris@maryland
davecl@orca.UUCP (Dave Clemans) (09/11/85)
Just a short note... There are at least a couple of other Unix mailers (other than sendmail/delivermail) that support aliasing and that have seen wide distribution. These are MH (from Rand & UCI) and MMDF (from CSNET, etc.) dgc
fair@ucbvax.ARPA (Erik E. Fair) (09/12/85)
In article <1722@orca.UUCP> davecl@orca.UUCP (Dave Clemans) writes: >Just a short note... > >There are at least a couple of other Unix mailers (other than >sendmail/delivermail) that support aliasing and that have seen >wide distribution. These are MH (from Rand & UCI) and MMDF >(from CSNET, etc.) My understanding is that while MMDF supports the sort of aliases that I meant, MH does not, as it is a user agent, rather than a system wide mailer ala delivermail or sendmail. Marshall Rose, if you're reading this, please feel free to correct me. Erik E. Fair ucbvax!fair fair@ucbarpa.BERKELEY.EDU
gds@mit-eddie.UUCP (Greg Skinner) (09/14/85)
> From: fair@ucbvax.ARPA (Erik E. Fair) > Unfortunately for this idea, all of the versions of UNIX, other than > the ones that come from Berkeley (2 & 4 BSD), have the most bletcherous > mailers that it would ever be your misfortune to encounter, which, > without exception, do not support the concept of an `alias' (e.g. > ucbvax!usenet, ucbvax!erik, and a few other things, can all point to > ucbvax!fair, without existing as entries in /etc/passwd). Ahem. In Unix System V, /bin/mail allows aliasing without passwd entries by placing in /usr/mail/user the line Forward to path!someone so that mail addressed to "usenet" can be forwarded to someone who can answer the questions. Unfortunately, it can't expand to more than one address. > So until AT&T and its imitators discover what *real* electronic mailing > systems are like (and what they do), we're stuck with the world as it is... Some machines in AT&T have sendmails running -- typically they're on the Datakit network. Don't malign AT&T mailers too much, they're not totally losing. One other thing, mailx(1) is a user agent, not a mailer. -- It's like a jungle sometimes, it makes me wonder how I keep from goin' under. Greg Skinner (gregbo) {decvax!genrad, allegra, ihnp4}!mit-eddie!gds gds@mit-eddie.mit.edu
stv@qantel.UUCP (Steve Vance@ex2499) (09/20/85)
In article <5307@mit-eddie.UUCP> gds@mit-eddie.UUCP (Greg Skinner) writes: >Ahem. > >In Unix System V, /bin/mail allows aliasing without passwd entries by >placing in /usr/mail/user the line > >Forward to path!someone > >so that mail addressed to "usenet" can be forwarded to someone who can >answer the questions. Unfortunately, it can't expand to more than one >address. Ahem. Ahem. Cough Cough. 1%% cat /usr/mail/bogus Forward to stv 2%% mail bogus Subject: test this is a test test. test. test. EOT Can't send to bogus "/q7/stv/dead.letter" 2/32 3%% This would only work if there was a line in my /etc/passwd file that started with "bogus". Also, you're right, there is no way to say "Forward to curly, larry, moe" or anything like that, let alone "Forward to ~/mailarchive" or "Forward to |lp". This limits the value of the "Forward to" feature, in my opinion. -- Steve Vance {dual,hplabs,ihnp4}!qantel!stv dual!qantel!stv@berkeley Qantel Corporation, Hayward, CA