% $Cambridge: hermes/doc/misc/crsids.tex,v 1.6 2009/05/26 14:03:45 fanf2 Exp $

\documentclass[a4paper]{article}

\usepackage{fullpage}
\usepackage{url}

\DeclareUrlCommand\email{%
  \def\UrlLeft{<}%
  \def\UrlRight{>}%
  \urlstyle{tt}%
}

\DeclareUrlCommand\domain{%
  \def\UrlLeft{@}%
  \urlstyle{tt}%
}

\setlength{\parindent}{0in}
\setlength{\parskip}{1em}

% rubric

\author{
  Tony~Finch \email{fanf2@cam.ac.uk}\\
  Mail Support\\
  University of Cambridge Computing Service\\
}

\title{
  CRSIDs and email addresses
}

\date{May 2009}

\begin{document}

\maketitle

%

\section{Introduction}

The Computing Service's Common Registration Scheme provides users
with identifiers that are recognized by many computer systems in the
University. CRSIDs are used as login identifiers and often as the
local part (before the ``@'') of email addresses. In the following
sections I list the advantages of constructing email addresses from
CRSIDs, especially in large mail domains like \domain{cam.ac.uk}. In
the final section I note some of their disadvantages, and how we
compensate by supporting ``friendly name'' addresses in department and
college mail domains.

%

\section{Advantages of CRSIDs}

Personal CRSIDs are formed from the user's initials with a serial
number appended. This construction has the following benefits:


\subsection{Stable}

CRSIDs are stable in two complementary ways.

Firstly, once a CRSID has been allocated to someone it is never
re-allocated to anyone else. This has a big privacy benefit: a
correctly addressed message sent to a CRSID email address will be
delivered to the person intended by the sender (or to no-one if the
recipient's account has been cancelled). It also has security
benefits, since web services on the Internet often send password
reminder emails, so control over an email address gives control
over those web accounts.

Secondly, a user's CRSID does not normally change. Even when a person
returns to the University we try to re-activate their old accounts,
and in several cases we have done so after absences of more than 20
years. (There are exceptions to this rule, for example a user can
change their CRSID when they change their name, or a returning user
may be allocated a new one if we fail to match their details with
their old account.)

Stability is especially important for email addresses that appear in
print, because of the long lifetime of books and papers.


\subsection{Fair}

The same CRSID allocation scheme applies equally to all users. This
helps with stability, since a user's email address does not change as
their seniority or affiliation changes. One CRSID is about as
desirable as another, so there is no arguing over which of two people
with the same name is more entitled to a particular email address.


\subsection{Scalable}

Although our active user population is about 40,000 the CRS has to
scale up to much larger populations because of the ``no
re-allocation'' stability rule.

The CRS has been operating for over 20 years, though the identifier
scheme dates back to the early 1970s. It comfortably copes with about
10,000 new identifiers every year, and a total of over 150,000
registered users.

The proportion of users with non-unique names rises as the population
increases. About 9\% of the active population has the same initials
and surname as another user. Some names are shared by more than a
dozen people. The proportion for all past and present users is over
18\%, with some names shared by over 50 people.


\subsection{Mnemonic}

It is reasonably easy to remember the CRSIDs of your most frequent
correspondents, because of the simple connection between their name
and their CRSID.

They are more helpful than phone numbers in this respect. They are also
more friendly than automated allocation schemes based on the user's
affiliation that some other universities use.


\subsection{Arbitrary}

The numeric part of a CRSID is obviously meaningless, which
discourages people from trying to guess email addresses and thereby
mis-address messages. In this way they are like phone numbers.

The arbitrariness is clearly visible. Sites with friendly-name email
address schemes must have rules to deal with name clashes, such as
adding middle initials. These lead to arbitrary variation in addresses
which is not obvious, and therefore makes the addresses less memorable
and more prone to incorrect guessing. At these sites it is common for
users with similar names to receive each others' email, which can lead
to embarrassment or worse.


\subsection{Automatic}

CRSIDs can be allocated automatically. This is absolutely crucial for
efficiently handling student admissions, when about 3500 accounts are
created at once.

Friendly name addresses often require manual adjustment. For example,
an automatic system would probably allocate me the ``friendly''
address \email{Frederick.Finch@ucs.cam.ac.uk} though I am usually
known as Tony. Similarly, users from the Far East often adopt European
names.


\subsection{International}

Email addresses can only use a limited character set, so it is not
possible to spell foreign names correctly when they contain accents or
are not written in the Roman alphabet. A set of initials is likely to
be less grating to the user than an incorrect full name.


\subsection{Un-gendered}

Some users prefer not to unnecessarily reveal their gender by
revealing their full name.


\subsection{Brief}

CRSIDs are short.


\subsection{Unobtrusive}

Most email software will show a correspondent's ``display name''
instead of their email address. Display names are for human
convenience; they do not need to be unique and can be
internationalized. Personal address books and institutional
directories allow users to look up email addresses by name.

(This is not really a benefit of CRSID email addresses, but more an
argument that tools exist to mitigate their disadvantages.)

%

\section{``Friendly name'' addresses}

The disadvantages of CRSID-based email addresses are:

\begin{itemize}

\item
They are cryptic, especially to those not in the know.

\item
They are difficult to communicate over the phone.

\item
The flat \domain{cam.ac.uk} domain gives no hints as to the user's
seniority or affiliation.

\end{itemize}

Because of this, the Computing Service encourages departments and
colleges to provide friendly name addresses in their mail domains.
Addresses like \email{Tony.Finch@ucs.cam.ac.uk} are somewhat analogous
to postal addresses in that they contain the user's name and location.
They also provide some indication of seniority since they are not
available to undergraduates, and often not postgraduates either.

Many users have multiple affiliations, typically a college and a
department. They can have multiple friendly addresses which all
deliver to the same inbox. They can send email using whichever address
they prefer or which is appropriate for each situation, be it college
or department business.

The Computing Service provides tools for colleges and departments to
manage mail domains that run on the CS's systems. These tools make it
easy for computer officers to set up personal email addresses for
their users that deliver to each user's \domain{cam.ac.uk} inbox.

College and department domains have much smaller user populations than
the central \domain{cam.ac.uk} domain. This greatly reduces the
likelihood that they will have to deal with name clashes. They can
also reduce the amount of churn by being restricted to staff, which
makes manual address allocation more feasible.

%

\end{document}

% eof