CERN Accelerating science

This website is no longer maintained. Its content may be obsolete. Please visit for current CERN information.

next up previous
Next: Central Mail Service: Status and Next Steps Up: cnl228.html Previous: The IT Training Service

Electronic Mail Addressing: Pitfalls and Inefficiencies

  Dietrich Wiegandt IT/DIS

PEMs and GEMs and their use at CERN

A growing number of electronic mail messages at CERN are failing because of invalid or "lazy" addresses. This article is meant to make you aware of possible misconceptions and imprecise information.

Typical mail addresses like jim@mymachine.moon.lh on messages travelling through the Internet are composed of a login_id (jim) and the name of a "physically existing computer" (mymachine.moon.lh), joined with an `@' sign. The resulting string uniquely defines a "mailbox", i.e. a file where the message should be stored.

In what follows this type of address will be called a Physical Electronic Mail address "PEM" (see note 1).

For quite a while we have been recommending the use of another type of mail address: the so-called Generic Electronic Mail address, "GEM". The format of a GEM is normally:

There is no name of a physical host between the character `@' and the "domain" which is the characteristic peculiarity qualifying a CERN mail address to be a GEM.

Attentive readers will notice that we have changed the definition of the acronym "PEM". It originally stood for Preferred Electronic Mail address and was used to allow people to specify on which of their computer accounts they wanted to receive mail - for most people it was a choice between IBM and VMS. Since we now recommend that you publicise your email address in the "GEM" format the word preferred has become confusing and so we have decided to abandon it.

Your GEM is a pointer to your registered PEM

When you use a GEM as an electronic mail address, what you specify is in fact a pointer to another address. The GEM points to the registered Physical Electronic Mail address (PEM) of the owner; you are using `indirect addressing'.

Be careful not to set your PEM to your GEM - this would cause a circular definition and invalidate your GEM.

There are several advantages to using a generic address:

Remember also that this mechanism does not imply that the registered PEM designates a mailbox on the MAILSERVer, i.e the recipient does not have to have an account on the MAILSERVer.

Problems with "lazy" addressing

The translation of a GEM to the corresponding PEM is usually done by a very efficient lookup in a hash table of registered PEMs using the GEM as a key. This lookup will obviously fail if the GEM is not an exact match of the key in the table, e.g. if instead of a fully typed firstname only an initial or no firstname at all is given in the destination address, for example instead of

In this case the system will invoke a "fuzzy match" program that does its best to find a match. Very often it will find more than one match so that the mail cannot be delivered and will be returned to the sender with a list of possible recipients, from which the sender must then choose the correct one by replacing the ambiguous destination address by an exact one from the list indicated in the non-delivery message.

Not only does this procedure use more resources than for an exact table match of the GEM, the saving of keystrokes for the original message is largely outweighed by the fact that the mail frequently has to be sent twice - sometimes even more often, because people tend to read the non-delivery mail only superficially (if at all). One of the best ways to waste mailservice resources is to make up a mailing list of GEMs where all firstnames are abbreviated to initials, so that the matching program has to be invoked for each and every recipient on this list. Please help us and yourself by using full GEMs.

A user may have more than one GEM

A registered PEM may be pointed to by more than one GEM, all belonging to the same person, but having the name part(s) spelled differently. Quite often a person is known by a firstname different from the "official" firstname printed in the passport and which was used for the registration at CERN. The "fuzzy match" program is unable to match "alternate versions" of firstnames, e.g. Dick to Richard, Ted to Edward, or Fritz to Friedrich.

Users known by "alternate" firstnames should make use of the possibility to have more than one valid GEM. Use the program "xuserinfo" (or its line mode version "cuserinfo") to register a preferred firstname and reset the first part of your GEM.

Service programs such as "xwho" will normally display a valid GEM for the user (or a PEM which is not pointing to a host at CERN), but other valid GEMs and the PEM will be shown when you select "valid mail addresses".

Common errors with addresses

  1. A very frequent misconception seems to be that once a user has got an account on the MAILSERVer he is reachable via an address of the form

    This is generally not true. Such an address is composed of the user part of a PEM and the host part of a GEM, so it is treated as a "lazy" GEM. Mail to such an address can only be delivered under exceptional circumstances, i.e. when the login_id matches exactly a family name which uniquely designates a person at CERN.

    Unfortunately, setting the parameter "user-domain" in the Pine configuration to "" can cause outgoing mail to have this invalid form as sender address. Leave that parameter empty, or, if and only if you have a MAILSERVer account, set user-domain to "" (no double-quotes in the Pine configuration file).

  3. Here is another invalid mixture of address formats:

    This address is invalid because it is using the name part of a GEM and the host-part of a PEM. Such an address will invariably fail, because Firstname.Familyname will be interpreted as a login_name.

A GEM as a forward address may cause bouncing mail

A GEM should be used only with extreme care as an address for mail forwarding, because it is too easy to cause mail to bounce between hosts. Keep in mind that your GEM is a pointer to your registered PEM, find out what that PEM is (use xuserinfo or cuserinfo) and what it would do if you used it directly instead of the GEM for forwarding.

Use of GEMs on outgoing mail

To propagate the use of GEMs on outgoing mail there is also a "reverse" translation, i.e. the translation of a PEM to the corresponding GEM. This translation is done when outgoing mail carries the registered PEM as sender address and passes through one of the MAILSERVer hosts, e.g. To avoid confusion about the "residence" of people this translation is not done when the host specified in the PEM is not located at CERN, nor is it done when the same PEM is used by several persons or services at CERN.

Where to ask for help

If you have any questions concerning mail addresses, even addresses not at CERN, contact, we may be able to help you.



  1. To register or reset your PEM you should use the program "xuserinfo" or its line mode version "cuserinfo" (entry Mail Info or Mail). For more information see the articles "EMDIR Stopped on September 16 1997" in this CNL and "How to Inspect and Update your Information in the Computer Centre Database (CCDB)" in CNL 225.

next up previous
Next: Central Mail Service: Status and Next Steps Up: cnl228.html Previous: The IT Training Service