Ralph Seichter via Postfix-users:
> * Geert Hendrickx via Postfix-users:
> 
> > The question was rather: shouldn't tools like "postmap", "newaliases",
> > etc, be a no-op on texthash: db's, rather than returning a fatal
> > error?
> 
> I prefer Postfix returning an error when asked to perform an operation
> which is not possible, nonsensical or otherwise indicates that there are
> misconfigurations or misunderstandings about Postfix's inner workings
> involved.
> 
> Errors are a way of telling users "don't do that" instead of asking
> Postfix to find its own interpretation of when it's OK to ignore or
> override what users tried to do. My two cents.

I think that there are different degrees of 'impossible', and that
they can correspond to different levels of severity.

Taking a 'regexp' or 'texthash' table as an example, a request to
add or delete a (key, value) should be a hard error because the
requested change will not be made. As the failure would affect the
result of future queries, a fatal error would be appropriate (non-zero
exit status, message on stderr etc.). This is existing behavior.

However, a 'create index' request would be harmless because the
table does not use a stored indexed file. As the failure would have
no effect on the result of future queries, a warning would be
appropriate (zero exit status, message on stderr etc.). This would
be new behavior; current behavior is to report a fatal error.

        Wietse
_______________________________________________
Postfix-users mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to