http://trac.roundcube.net/ticket/1486386
What do you think about that. We're storing username and mail_host in users table and then checking both. This is problematic. I assume that it's done this way because in some environments username is a login not email address, but maybe (in rcube_user::query) instead of
SELECT * FROM users WHERE mail_host=? AND username=?
we should use
SELECT * FROM users WHERE username=?
when username is an email address.
I think it would work better for us if there was only one user record per actual user, but then any number of mail accounts associated with that user record. However I suspect that your current setup also works well for others who are not in the same setup as a university, but instead they are maybe a hosting provider (because then they don't need to worry about name space collisions).
Todd
A.L.E.C wrote:
http://trac.roundcube.net/ticket/1486386
What do you think about that. We're storing username and mail_host in users table and then checking both. This is problematic. I assume that it's done this way because in some environments username is a login not email address, but maybe (in rcube_user::query) instead of
SELECT * FROM users WHERE mail_host=? AND username=?
we should use
SELECT * FROM users WHERE username=?
when username is an email address.
I work for an ISP and recently bumped into this while migrating one of our webmail frontend boxes. Turns out, that if we decide to change the backend mail server hostname, we also have to do it in the db tables for the webmail client, otherwise all users would get new profiles when logging in.
Rūdis
On Fri, 2010-03-12 at 09:05 -0600, Todd Trann wrote:
I think it would work better for us if there was only one user record per actual user, but then any number of mail accounts associated with that user record. However I suspect that your current setup also works well for others who are not in the same setup as a university, but instead they are maybe a hosting provider (because then they don't need to worry about name space collisions).
Todd
A.L.E.C wrote:
http://trac.roundcube.net/ticket/1486386
What do you think about that. We're storing username and mail_host in users table and then checking both. This is problematic. I assume that it's done this way because in some environments username is a login not email address, but maybe (in rcube_user::query) instead of
SELECT * FROM users WHERE mail_host=? AND username=?
we should use
SELECT * FROM users WHERE username=?
when username is an email address.
List info: http://lists.roundcube.net/dev/
http://trac.roundcube.net/ticket/1486386
What do you think about that. We're storing username and mail_host in users table and then checking both. This is problematic. I assume that it's done this way because in some environments username is a login not email address, but maybe (in rcube_user::query) instead of
SELECT * FROM users WHERE mail_host=? AND username=?
we should use
SELECT * FROM users WHERE username=?
when username is an email address.
Even when username isnt an email address it's sometimes useful. We have a large internal imap cluster, and sometimes we experiment with new versions or whatever, and right now thats difficult because if we change the imap server in roundcube all user settings are gone. In situations like ours, 1 ISP and users can not pick other servers themselves anyways, it makes no sense to check mail_host.
Cor
List info: http://lists.roundcube.net/dev/
-
A.L.E.C -
Cor Bosman -
Rudolfs Osins -
Todd Trann