db, field length too short

[runout-at]

The field length if pop and smtp fields in the users table is too short.
It's only 255 bytes which is very less on long user/domain names.
I would suggest 4096 bytes for those fields. AFAIR 4096 is the spec for path length in the file system.

[Udera]

username doesn't have to be a mail address, so we could allow larger usernames. pop and smtp don't need to be mail addresses (can be pipe commands), so could be longer as well.

mail addresses have a maximum length of 254 characters:
https://en.wikipedia.org/wiki/Email_address#Syntax
https://tools.ietf.org/html/rfc3696#page-6
we have currently localpart=192 (should be 64) and domain=64 (should be 255)
but the total length must not exceed 254 characters (we do not verify that at the moment):
http://www.rfc-editor.org/errata_search.php?rfc=3696&eid=1690

[runout-at]

yes, but the fields contain the whole path like
/var/opt/vmail/mailserver.my-extra-long-domain-name.at/my-extra-long-username/Maildir
and as you stated maybe a pipe command which can also be longer than 255 char.

i run into that problem by upgrading every time because i already have bigger fields in my database. and the update script shortens the fields.

[rimas-kudelis]

Let's patch this!
We probably have to adjust maximum username length for #106 as well.

[Udera]

I added an address-check: Udera@3ab9e40

@rimas-kudelis: Can you fix the db-fields? This would be part of your db-migration-script.

[rimas-kudelis]

Should we shorten localpart as well? It doesn't make a lot of sense to have a localpart longer than 60 characters, but it wouldn't be nice to cut it off silently as well. If we do shorten it, I think it should be mentioned in release notes or upgrade notes.

[Udera]

We should shorten localpart and put this in the release notes. I thought
about a small tool to check the existing data base if all users have
valid addresses. But testing my address validation, I considered that
you would have to use extremely long mail addresses. That's very
unlikely an issue for anybody.

On 2016-02-22 7:23, Rimas Kudelis wrote:

Should we shorten localpart as well? It doesn't make a lot of sense to
have a localpart longer than 60 characters, but it wouldn't be nice to
cut it off silently as well. If we do shorten it, I think it should be
mentioned in release notes or upgrade notes.

Reply to this email directly or view it on GitHub [1].

Links:

[1] #108 (comment)

[Udera]

@rimas-kudelis : Did you update the length of the database-fields? I didn't touch that because you already worked on a few changes in here: #72

[rimas-kudelis]

No, not yet. Feel free to take over if you want to do this quicker than myself. :)