What characters should I use or not use in usernames on Linux?

Question

I want to create user accounts named after a domain name. adduser complains that the usernames need to match the NAME_REGEX regular expression.

adduser: Please enter a username matching the regular expression configured
via the NAME_REGEX configuration variable.  Use the `--force-badname'
option to relax this check or reconfigure NAME_REGEX.

I can add the users using useradd without complaint. Is there a reason that I shouldn't modify the regular expression to allow ., - and _?

What characters will cause problems and shouldn't be allowed in usernames?

This is the default NAME_REGEX.

NAME_REGEX="^[a-z][-a-z0-9]*\$"

Answer 1

My advice to you is to follow the standard recommended by the default NAME_REGEX. You can actually put nearly anything in a user name under *NIX but you may encounter odd problems with library code that makes assumptions. Case in point:

http://blog.endpoint.com/2008/08/on-valid-unix-usernames-and-ones-sanity.html

My question to you: do you have a lot of domain names that would collide with each other if you stripped out the unusual punctuation? For example, do you have both "QUALITY-ASSURANCE" and QUALITYASSURANCE" as domain names? If not, you could simply adopt a policy of stripping out the unusual characters and using what's left as the user name.

Also, you could use the "real name" section of the GECOS field in the /etc/passwd information to store the original, unmodified domaain name, and scripts could extract it pretty easily.

Answer 2

From the NAME_REGEX can be deduced that everything but a through z in upper- and lowercase and the number 0 through 9 would be bad.