Re: [Exim] Sendmail compatibility

Página superior
Eliminar este mensaje
Responder a este mensaje
Autor: Jeffrey Goldberg
Fecha:  
A: alexpro
Cc: exim-users
Asunto: Re: [Exim] Sendmail compatibility
On Tue, 6 Mar 2001, Alexey Promokhov wrote:

> On Mon, 5 Mar 2001, Jeffrey Goldberg wrote:


> > It appears that there is an undocumented feature of sendmail which treats
> > the particular command-line syntax error as an implicit use of "-t".
> > Should that "feature" be added to exim for compatibility?


> Yeap. Or Exim should not be claimed as 'Sendmail interface compatible'. Or
> this stuff should be added to documentation or FAQ.


It is hard to maintain interface compatibility with undocumented
features. Using sendmail as you (or FidoGate) used it really is an error.
So it is not clear to me whether exim should try to follow all sendmail
undocument features.

Keep in mind that the undocument behaviour of sendmail (or actually
behaviour which runs counter to the documentation) may vary from
implementation to implementation. Look at the discussion of -t in the
"Calling Interface" section of the exim specification.

Another place where sendmail doesn't follow its document (but where exim
does) is what happens in these cases where the input begins with a blank
line. Some (all?) versions of sendmail just ignore the leading blanks and
takes the initial set of non-blank lines as headers. Exim follows
sendmail's documentation and treats those as headerless inputs.

I suppose that these various sendmail undocumented peculiarities ought to
be discussed in the exim documentation.

But really, people shouldn't be relying on undocument peculiarities of
sendmail in the first place. FidoGate simply has it wrong. (Although I
was the one who pointed out the sendmail behavior that eventually lead to
the creation of the extract_addresses_remove_arguments option in exim.)

> FidoGate is not wrong, but its documentation does not say something about
> this.


Calling sendmail as

sendmail -f some@address

is broken according to sendmail documentation as I recall it, and
according to the sendmail logs you posted. The fact that that broken way
of calling sendmail actually works with (some versions of) sendmail is a
surprise.

> It is not problem to add -t option to sendmail, it's not obivious.


It should have always been there.

-j

--
Jeffrey Goldberg
I have recently moved, see http://www.goldmark.org/jeff/contact.html
Relativism is the triumph of authority over truth, convention over justice