wyle@ethz.UUCP (Mitchell Wyle) (08/24/87)
Why I don't use ean Mitchell Wyle wyle@ethz.uucp The University of British Columbia's X.400 mail system "ean" is the electronic mail system of choice for the ETH computer science department. I have used the system for over a year; it is the system with which I am most experienced, but I have discovered some features in the system which have led me to the decision to change mail systems. Most major decisions of this nature are prompted by unfortunate accidents. Such is the case in this situation. Because ean uses a bizarre method of storing its data (db5), one must periodically "clean" the data files ean creates. Otherwise, they would continue to grow without bound, until the disk is full. One day, the "eanrebuild" command destroyed all of my ean data. Upon restoring the data from a backup tape, ean refused to read the restored data. A mail system which 1) needs a dangerous, periodic "clean-up" command, and 2) cannot read restored data is unacceptable. The data cannot be moved to tape and back to disk. No backup is possible; therefore ean is unacceptable. The eanrebuild command is simply a nuisance which does not belong in a computer system. One could live with such an inconvenience if backup were possible. Ean's user interface is as poor as most other electronic mail systems'. There is no way to specify rules and actions (it is not programmable). It has a flat storage method (one level of folders). It does not take advantage of a video screen terminal. It has no information retrieval system for mail messages. Ean cannot run on all unix systems. It is specific for the DEC vax. It's files cannot be used by other applications. Almost all ean messages I send arrive with an annoying "Parse error of original messages in Switzerland" error message, which leads me to assume that the lower layers of the system are as weak as the user interface and no-backup features. The public domain, University of California Berkeley's (UCB) Unix mail system is not much better, but at least there is no "rebuild" command, and the files are standard unix text files. One can use the Unix shell environment, the "mail-tool" interface on Suns, and the file directory tree to add structure to mail messages. UCB Unix mail is programmable and richer in commands. It is standard on many different machines, and identical on Suns and vaxen. Most importantly, the data are interchangeable, and can be backed up! Ean? No thank you. -- Mitchell F. Wyle | csnet or arpa: wyle%ifi.ethz.ch@relay.cs.net Instituet fuer Informatik | uucp: wyle@ethz.uucp ETH Zentrum / SOT | Telephone: 011 41 1 256 5237 8092 Zuerich, Switzerland
taylor@hplabsz.HPL.HP.COM (Dave Taylor) (08/25/87)
Mitchell Wyle writes an interesting article about why he doesn't use the EAN X.400 mail system from the University of British Columbia in this group. I'd like to discuss some of the points he raises a bit further... First off, I think it is important to note that X.400 and the related ISO/CCITT standards are indeed an up and coming standard for the interchange of electronic mail in the world (certainly in Europe). In fact sources from both DEC, HP, and some third party market analysts tell me that it is almost impossible to sell any large computer boxes to firms there unless they have available an X.400 mail system... In an environment of that nature, it is somewhat ironic to read that a sophisticated, knowledgeable Unix user like Mitchell decides to use something as simple as the Berkeley Mail system instead of the EAN package. My first curious question is how long will Mitchell be able to use Berkeley Mail to send and receive mail there in Switzerland? I can imagine in the very near future his machine suddenly not speaking the "RFC-822 style" mail that Berkeley Mail expects and generates. Or, as an alternative, having the extra overhead of a gateway/translator for each message sent, even those local to his machine. Note that I haven't made any comments about the relative merits of the X.400 protocol suite/standard nor of the RFC-822 (and related documents) standard either. The tough fact of the matter is that it is almost completely irrelevant whether we like the standard or not - X.400 is happening now regardless. There are already countries where, as I said previously, you *cannot* sell machines unless you offer a complete X.400 system. (interestingly, the best selling X.400 package, All-In-One from Digital Equipment Corporation, currently has a number of blatant protocol violations on its lower-level interactions with other X.400 systems (but we can be sure that DEC is working like mad to fix that as soon as they possibly can!)) This is not to make light of the problems of an electronic mail package user interface! From dint of the experiences with not only the Msg and Elm mail packages, but from the research I've put in on various articles on electronic mail systems (the latest being in the book "The Unix Papers" from Howard and Sams Publishing...) I've found that perhaps the *most* important aspect of any system, and especially a system like one for the creation, transmission and reception of electronic mail, is in fact the user interface. So with that preamble, I'd like to highlight the deficient areas of the EAN package, based *purely* on what Mitchell Wyle posted in his article. The first problem that is raised is that of the data storage method. Not only does EAN use a unique format for the storage of data (which, I would venture a guess, is related to X.408 (MHS: Encoded Information Type Conversion Rules) and X.409 (MHS: Presentation Transfer Syntax and Notation) standards if nothing else..) and that a side effect of this is that 1. the folder space is one-level deep (e.g. there is no capability for having folders within other folders), 2. the utilities that are available for keeping your dataspace clean aren't totally reliable (e.g. "eanrebuild"), and 3. EAN didn't like the data that was restored from a backup tape. One of the stranger aspects of the X.400 suite from our anglocentric point of view here in the US is that it doesn't store everything in an English-based format. In fact, the method by which the message envelope (as they say) is built up and saved is designed to be as *in*dependant of host language as possible. One of the 'headers' of an X.400 message is in fact the language that it was composed in, so that the displaying system can easily choose the appropriate character set. An obvious side effect of this is that the data storage format for the messages cannot suddenly be in a different, '822 style, format. So instead, you're left with the necessity of having to use various utilities to help maintain your data space. Given that, I would suspect that the EAN group would be more than just interested in hearing about problems with the reliability of their system! Certainly if you can fully describe the situation whereby the "eanrebuild" program destroyed your saved messages it would at least be a big help in ensuring that they can fix it and make sure it doesn't happen to other people! And as to the problem with data off of a backup tape, I suspect that it is a subtle problem with file modification times or something, and that, again, if you were to *tell* the EAN people about the situation they would quickly remedy it. [general philosophical comment here: I find it unfortunate that so many people aren't willing to work with their software vendors in situations such as this - software is just like any other complex system and it occasionally needs some fine tuning. But there is *no* way for the intrepid vendor to achieve this if the users have something go wrong then simple stop using the product, instead of letting 'em know about it! Let's get on the ball, people!] Mitchell then goes on to talk about some aspects of the User Interface to the EAN X.400 mail system. Specifically, he states that a couple of the problems he has are that; 1. there is no way to specify rules and actions and otherwise make it a programmable interface, 2. there is no way to have multi-level folders, 3. it doesn't exploit the additional resource of the CRT *screen*, and 4. it has no intelligent retrieval system for archived messages. I must admit that I'm not sure exactly what Mitchell is talking about in his first point here - I have on my machine a program which 'screens' my electronic mail before allowing it to be delivered to me, and it can reply to mail automatically, store it in folders for later perusal, or even instantly display it to the screen. But that is one of only three systems that I know of that offer this sort of functionality and none of them are in widespread use (they are "filter", from the Elm Mail System, "mh-hook" from the MH package, and MEP from John Antypas). I'm very interested in hearing more about what is being suggested here (and shan't bias it by talking about how *I* think mail systems should work!) The problem with multi-level folders, and the related problem of not having an intelligent retrieval system are actually symptomatic of a much larger problem with generalized information retrieval systems. I believe that the *real* answer to this is to have a hypertext system available and to obsolete the entire idea of folders. I know of a couple of systems in the works that support just this, and I imagine now that Apple has released HyperCard for the Macintosh that it's just a matter of time before some bright cookie exploits that for the same area. The idea here would be to specify a list of criteria to base selection of possible messages on - sort of like full-text keywording - and then have the system select messages based on that. (I think this is really the wave of the future and I'm looking forward to having it on MY machine soon!) As to the EAN interface not exploiting the resources and space of the CRT screen, well, all I can say is that Andrew Draskoy of the University of British Columbia and I have talked a couple of times about full-screen interfaces for the EAN system and that he has recently requested a copy of the Elm Mail System from me for some ideas of how to exploit a full screen and still retain the ease of use of a simple system...I wouldn't be suprised to see them release an impressive and powerful new interface to their system within a year. By the way, Mitchell also refers to the Berkeley Mail package as being "in the public domain". This, of course, is not the case at all, and the package is in fact covered under the same copyright and licensing restrictions as the rest of the BSD binaries and sources. Finally, I hope those of you that have managed to read this far have found this an informative note. I would like to hear further discussion on electronic mail user interfaces and shall be forwarding the comments as appropriate to interested other parties. (there are also a number of mailing lists dealing with mail user agents, mail handling systems, the X.400 protocol suite, and so on - check the "list of lists" that sporadically appears on the net...) From the far corners of the globe, -- Dave Taylor Hewlett-Packard Labs
daveb@geac.UUCP (Brown) (08/26/87)
Mitchell Wyle comments: | .... Because ean uses a | bizarre method of storing its data (db5), one must periodically "clean" | the data files ean creates. Otherwise, they would continue to grow | without bound, until the disk is full. One day, the "eanrebuild" | command destroyed all of my ean data... Dave Taylor replies: | The first problem that is raised is that of the data storage method. Not | only does EAN use a unique format for the storage of data ... | An obvious side effect of this is that the data storage format | for the messages cannot suddenly be in a different, '822 style, format. | So instead, you're left with the necessity of having to use various | utilities to help maintain your data space. I think that dave has a misconception here: X.400 defines a format for the contents of a file which constitutes {message*}. This format is very non-ascii and not fully convertable to RFC 822 ascii. It does not, as I read the specification (which is not fully convertable to a human language), require the storage of the messages in a file or file sequence that is not manipulatable with more-or-less normal system tools. The use of db5, if true, is a decisision made by the ean developers, and appears to have caused a problem... --dave (i think i like the format, but i can't read the spec) cb -- David Collier-Brown. {mnetor|yetti|utgpu}!geac!daveb Geac Computers International Inc., | Computer Science loses its 350 Steelcase Road,Markham, Ontario, | memory (if not its mind) CANADA, L3R 1B3 (416) 475-0525 x3279 | every 6 months.
lamy@utegc.UUCP (08/27/87)
In article <186@bernina.UUCP> wyle@ethz.UUCP (Mitchell Wyle) writes: >The public domain, University of California Berkeley's (UCB) Unix mail >system is not much better, but at least there is no "rebuild" command, >and the files are standard unix text files. One can use the Unix shell >environment, the "mail-tool" interface on Suns, and the file directory >tree to add structure to mail messages. UCB Unix mail is programmable >and richer in commands. It is standard on many different machines, and >identical on Suns and vaxen. Most importantly, the data are >interchangeable, and can be backed up! UCB mail is not public domain, and it does not really mesh well with the Unix directories (retrieving an individual message is a pain). There is no programmability either. The mailtool on the Suns does not allow you to chose which editor you want to edit the messages. UCB mail is not screen oriented. In fact, I must admit that I really don't give much credence to your claim that UCB mail's interface is any richer than EAN's. Yes, I have used both. ELM will give you a screen oriented interface. If you want full configurability, then use MH, from within an editor if you wish. MH uses the directory structure for folders and gives you a query facility to retrieve messages. In fact, there is no real reason either why you can't have the best of all worlds. This message is composed in a MH mode in GNU-emacs (with roughly the same features as mailtool), will be processed by the standard Unix sendmail, and may end up being sent by EAN, acting as a message transfer agent (we also have connection to the Internet, UUCP and Bitnet). Getting EAN to interface with other mailers can be a pain, but there is hope (this site is a living proof that an interface with the standard Unix setup can be done, and the VMS/EAN gatewaying seems to work fine as well), In other words, a little patience may be in order. Wrt your comment about parse errors, note that even relay.cs.net will complain about (hackish) addresses with a % in them. I don't see this as a problem with EAN. I am not the greatest fan of X.400 mail (especially with respect to the proposed organization-oriented directory service), but I don't think EAN is such a bad deal for a site with a single connection to the outside world. Jean-Francois Lamy lamy@ai.toronto.edu (CSnet,UUCP,Bitnet) AI Group, Dept of Computer Science lamy@ai.toronto.cdn (EAN X.400) University of Toronto, Canada M5S 1A4 {seismo,watmath}!ai.toronto.edu!lamy
zentrale@rmi.UUCP (08/28/87)
In article <719@hplabsz.HPL.HP.COM> taylor@hplabsz.UUCP (Dave Taylor) writes:
: First off, I think it is important to note that X.400 and the related
: ISO/CCITT standards are indeed an up and coming standard for the
: interchange of electronic mail in the world (certainly in Europe). In
: fact sources from both DEC, HP, and some third party market analysts
: tell me that it is almost impossible to sell any large computer boxes
: to firms there unless they have available an X.400 mail system...
Espacially interesting is: X.400 is the prot of the administrations. Those
"official" Mailboxsystems use X.400 and will gateway to others. If a host
wants to paricipate the commercial networks beeing established, it has to
to it by using X.400.
:
: One of the stranger aspects of the X.400 suite from our anglocentric
: point of view here in the US is that it doesn't store everything in an
: English-based format. [...]
What means "stranger aspects? There are countries with users, who don't
speak English . So an language indepent Prot has its merits.
:
: [general philosophical comment here: I find it unfortunate that so many
: people aren't willing to work with their software vendors in situations
: such as this - software is just like any other complex system and it
: occasionally needs some fine tuning. But there is *no* way for the
: intrepid vendor to achieve this if the users have something go wrong
: then simple stop using the product, instead of letting 'em know about
: it! Let's get on the ball, people!]
:
Stands for itself, just repeated to be kept in mind!
:
: -- Dave Taylor
: Hewlett-Packard Labs
Trying to get hand on X.400 Software for UN*X I was given the address
of Sidney International. Contacting them resulted in the following
information: Source License Pound Sterling 250,000, Binary 5,000! Just
for an mailing system ... :-)
Rupert Mohr
RMI Aachen
peter@ethz.UUCP (Peter Beadle) (08/31/87)
In article <719@hplabsz.HPL.HP.COM> taylor@hplabsz.UUCP (Dave Taylor) writes: >First off, I think it is important to note that X.400 and the related >ISO/CCITT standards are indeed an up and coming standard for the >interchange of electronic mail in the world (certainly in Europe). In >fact sources from both DEC, HP, and some third party market analysts >tell me that it is almost impossible to sell any large computer boxes >to firms there unless they have available an X.400 mail system... To this I agree, the X.400 protocols are quiet nice. Unfortunately the group here responsible for maintaining the ean mail system repeatedly tell me "the EAN implementation of the x.400 protocol package is a heap of sh*t". I manage the mail and wide area network for the machine Mitchel Wyle's mail goes through. This machine has 4 gateways to disparate networks (uucp, cipon, ean and ACSnet) and sometimes (depending on the weather) acts as the ean-csnet gateway. ACSnet runs itself and tells me what it's been doing once a month (it suports remote job execution, binary transfer (on 7bit, 8bit and x25 lines), file server, remote printing, and acts as an inter network gateway) all the others require work. All the machines in a cluster of ~40 run uucp (over ethernet) as the transport level with the upas mail system (Bell Labs) doing the routing and acting as a name server. Running on top of that we have a modified /bin/mail, ucb/mail, mh and sun's mailtool interface to /ucb/mail there are also a gagle of emacs based mail programs and Mail (if anyone remembers that antique). On the gateway machine we also have ean running. I don't administer ean and frankly wish it would go away. I have used the system and gave up after many complaints from correspondants who got tired of "auto forwarding" as the non-delivery reason in the bounce messages they received. It also appears that there is no way in ean to do any of the usefull mail filtering things that make email usefull in administering machines. I say apparently because the only documentation available is the online help from inside the ean system, there is no machine manual and no paper manual available to users of the system and I doubt it is available to the administrators. >package. My first curious question is how long will Mitchell be able >to use Berkeley Mail to send and receive mail there in Switzerland? I >can imagine in the very near future his machine suddenly not speaking >the "RFC-822 style" mail that Berkeley Mail expects and generates. Or, >as an alternative, having the extra overhead of a gateway/translator >for each message sent, even those local to his machine. Yes the ean-isation of Switzerland is well underway. It's a pity they couldnt be bothered to develope their own x400 system instead of importing a bad one from Canada. I am also not really worried about the translation time, I simply want a mail system that works. Most sites understand either rfc-822 or uucp mail headers, we provide both. We have no trouble gatewaying into ean with our rfc-822 headers. Unfortunately the ean system can't even provide a from address to our mailer so it can correctly set the from fields on mail coming from ean into uucp/rfc-822. I have also been told that on the next release of ean they will provide a date field in the header. A great leap forward. >standard either. The tough fact of the matter is that it is almost >completely irrelevant whether we like the standard or not - X.400 is >happening now regardless. There are already countries where, as I said >previously, you *cannot* sell machines unless you offer a complete X.400 The point is the ean implementation of x.400, not x.400 itself. Could you please supply some proof for your claim about x400 and computer sales. >So with that preamble, I'd like to highlight the deficient areas of the >EAN package, based *purely* on what Mitchell Wyle posted in his article. Have you ever tried to use the ean package. If not I sugest you try to get access to an ean system and see how it feels. (give me a call and I will give you our modem number). >The first problem that is raised is that of the data storage method. Not >only does EAN use a unique format for the storage of data (which, I would >venture a guess, is related to X.408 (MHS: Encoded Information Type >Conversion Rules) and X.409 (MHS: Presentation Transfer Syntax and Notation) >standards if nothing else..) No it's because ean seems to use the cheap nasty dbm data base that comes with 4.2 with the result that the files it creates can't be manipulated with standard unix comands. Still I supose popple live with the Itallian postal service (average delivery speed 6 m.p.h. and a nasty habit of shreading the mail backlog. >One of the stranger aspects of the X.400 suite from our anglocentric >point of view here in the US is that it doesn't store everything in an >English-based format. In fact, the method by which the message >envelope (as they say) is built up and saved is designed to be as >*in*dependant of host language as possible. One of the 'headers' of >an X.400 message is in fact the language that it was composed in, so >that the displaying system can easily choose the appropriate character >set. An obvious side effect of this is that the data storage format >for the messages cannot suddenly be in a different, '822 style, format. >So instead, you're left with the necessity of having to use various >utilities to help maintain your data space. All very interesting, if you could rely on your data base the it might even work. If you could rely on your communication medium there would be no need for uuencode/decode, binhex and innumerable other programs that convert binary into ascii and back and you would be much further towards a usable "communications" system. I think the reliable communication system would bring you a lot more than selecting the language style in the mail header. >Given that, I would suspect that the EAN group would be more than just >interested in hearing about problems with the reliability of their >system! They already know of the problems, see my earlier quote about their summary of the system. > people aren't willing to work with their software vendors in situations > such as this It's a little hard when the software vendor is a University on the other side of the Atlantic and the people who developed th code have long since graduated and moved on. I refer you to another article in this news group about a commercial x400 system, source lisence only 250,000 pounds sterling. Per cpu lisence only 5000 pounds. Summary: There is nothing wrong with the x.400 protocols. It would be good if they become a european standard. The currently available x-400 implementations stink. I would love to hear somone from the University of British Columbia defend their system. Ean appears to lack: . adequate documentation for users, . support for mail filtering, . methods of turning off stupidities (like auto forwarding as a non delivery reason) . alternate presentation system (you get the ean presentation system and thats all) . support for file transfer and remote execution which where there in the most primative uucp mail systems Peter Beadle uucp peter@eiger.uucp ean/csnet peter@ifi.ethz.ch (try this one for a non delivery bounce message) uucp mcvax!cernvax!ethz!eiger!peter acsnet peter@bernina.ch
lubich@ethz.UUCP (Hannes Lubich) (08/31/87)
Peter, just one point: I don't think that I will try to defend the old ean version that we are running here (got, that, John :-)) but I would wish that everybody would document software in the way ubc did it for ean. The user documentation is very good (there is a seperate documentation for VMS managers, Unix- managers and simple users, there are least two quick reference cards and we added a whole bunch of additonal information (including a telephone hotline service and the like). In addition, a large ammount of documentation is available on an online basis, so there is really no reason for complaining about the documentation. We all agree that EAN V1 has some major disadvantages but I think that all the people that now tend to flame ean should have a) spoken up earlier b) spoken up in a more usable fashion c) directed their complaints to ubc --Hannes -- ~ UUCP/Usenet : {known world}!seismo!mcvax!cernvax!ethz!lubich ~ CSNET : lubich%ifi.ethz.chunet@relay.cs.net ~ ARPA : lubich%ifi.ethz.chunet@csnet-relay.arpa The usual disclaimer : No, it wasn't me, somebody must have used my account.
taylor@hplabsz.HPL.HP.COM (Dave Taylor) (09/03/87)
Rupert Mohr replies to my comments about EAN and X.400 with some interesting comments of his own, but at one place he misses my point entirely, alas: >: One of the stranger aspects of the X.400 suite from our anglocentric >: point of view here in the US is that it doesn't store everything in an >: English-based format. [...] > >What means "stranger aspects? There are countries with users, who don't >speak English . So an language indepent Prot[ocol spec] has its merits. I think the problem here is with the word 'anglocentric'. I used it to mean 'centered on the anglo/english-speaking world'. The comment was more to the effect of 'it is unfortunate that here in the US we're so convinced in the ultimate value of English that we, for the most part, view any system that is to support other languages by *not* being in English as suspicious' (althought that isn't much clearer either!!) In general I think that X.400 is a good thing (are we going to start talking about the value of X.400 here??) and that since it specifies a non-ascii-based message envelope it makes the most sense for the messages to also be stored in that format. (a strong case can be made that it should be translated to the local language upon storage, though, and I'd be interested in hearing peoples ideas on this either way...) As I've tried to say too, however, the Unix system especially is very much oriented around having 'flat' ASCII files and there are a plethora of useful and interesting tools that you can then use to wreak havoc on these text files. Storing X.400 mailboxes as non-ASCII data unfortunately loses this major Unix advantage, and the X.400 vendor is then put in the position of offering their own tools to replace and/or supplement the already-built-and-tested Unix ASCII tools... six of one, half-dozen of the other.... -- Dave Taylor Hewlett-Packard Labs