Stephen Fischer wrote:
> I would prefer not to have sendmail do it, because it's a larger drain on the
> resources and because it's not really sendmail's job to handle these things
> (as I understand it). The reason for this is because at times the process
> that informs us about new users takes some time, so the users may be
> distributing the mail address before we are told about it. As a result, we
> could be bouncing important mail, which is not politically popular.
I agree that it is not sendmail's job. If you are just dealing with one
domain name, it makes sense to not to use sendmail. However if
you have the scenario where you have to support say xyz.com and
xyz.net (which is the case by and large for most companies these
days), which requires name mapping, I don't know where else but
sendmail, you could do what you are trying to do!
>
> The resource and security issues don't need to be a problem because someone
> could just mailbomb the valid mail addresses. And, hey, I'm not administering
> the system, so if they want to do it that way, that's their decision.
Mail bomb just affects the targeted account in a quota enabled system.
Where as the system you described, could potentially bog down entire
mail system, affecting everybody.
>
> The main problem with what you suggest is that you have the overhead of
> checking the existence of every mailbox before delivering the item, which is
> substantial, and why it would be optimal if deliver simply had a flag that
> created any non-existing mailboxes.
>
> Stephen
There is always a lookup overhead, it may be low towards the delivery end
and may make sense to implement it there, provided you have no name mapping
requirements.
__
Seva
>
> quoth Seva Adari:
> | I have not done anything like what you are suggesting, but
> | from what I know about sendmail, it ought to be possible to
> | create automatic mailboxes. Requires some coding though!
> |
> | I am little curious to find out why would you want to be
> | doing such a thing though! It can get you into resource
> | related troubles (not to mention security nightmares), given
> | that every mailbox requires that you set aside some space.
> |
> | I would look at the part where sendmail provides a hook for
> | mapping the mail ids via "virtusertable". Under sendmail there
> | is a file called "map.c" where they have the functions for
> | mapping the in coming mail id to internal id. Along lines of
> | exiting functions you could create a new one or override an
> | exiting function and check for existence of current mail id
> | in the system and if none found create a mailbox and then get
> | back into sendmail process flow.
> |
> | __
> | Seva
> |
> | Stephen Fischer wrote:
> |
> | > I'm interested in having the deliver program automatically create mailboxes
> | > when it receives a mail item addressed to an address of the correct format but
> | > which does not already exist. I found some mail in the archive from people
> | > asking a similar question but no answers. Is this capability there? If not,
> | > has anyone implemented something similar?
> | >
> | > Software: cyrus imapd 1.6.24, cyrus-sasl-1.5.20/24, sendmail.8.10.1, solaris 8
> | >
> | > Format (for reference): [a-z][0-9]{7}
> | >
> | > Basically, what we would like to have happen is that the deliver program
> | > checks the format of the address. If it is invalid, then the message is
> | > rejected. If it is valid, then it is delivered, creating the mailbox if it
> | > does not already exist. The first half is easy enough to do with a wrapper
> | > program that then calls deliver, but the second half seems to be best handled
> | > by deliver itself.
> | >
> | > Thanks for your help/thoughts,
> | > Stephen
> |
