[Pkg-mailman-hackers] Bug#244673: Your Mailman Debian bug

[Bernhard Reiter]

On Mon, Nov 28, 2005 at 09:46:22PM +0100, Lionel Elie Mamane wrote:
> > http://ftp.intevation.de/users/bernhard/mailman/mailman-2.1.4-avoid-headerfolding-python21.diff

I am just comparing it to 77_header_folding_in_attachments.dpatch.

> I took a look at the patch. In the class "Generator" you define, you
> completely ignore the argument mangle_from to __init__. If it is not
> used, I would expect it not to be taken as an argument of __init__ at
> all. 

I accepted it as part of the promissed interface.
Note that you, in replacing maxheaderlen, will not fullfill
the promise of Generator __init_ interface anymore.
It should not be of practical relevance in this case, though.

Note also that clone() now also does something different in
a subtile way in your patch and depends on a Python version >2.2.2.
My patch worked for versions 2.1.x, too.
No problem for Debian of course.

> Or, on the contrary, to be used and passed to the
> email.Generator.Generator.__init__ call. It is probably best to pass
> it to email.Generator.Generator.__init__, and set it to "False" in all
> instances where one writes to a one-message file or to a string
> (nearly everywhere) and to "True" when one writes to a mailbox (in
> Mailman/Mailbox.py). 

I agree that it would be better 
to only have the mangeling in the case of writing an mbox file.
How to reach that goal is a matter of coding style,
my change was a bit easier to track for me back then I guess.

> Current OpenPGP implementations do
> 'From '-mangling themselves anyway before clearsigning. I dunno
> whether S/MIME implementations do, too.

I do not know for sure either.
I would assume that any reasonable
OpenPGP or S/MIME implementation would not mangle the from before signing.
Otherwise it would be a "modification" of the message
and the email would contain something that the user did not write.
It is a minor modification, but still it is a modification.
And maildir system can save a "From " in the body without problems
and SMTP also has no problem.

> Same story for the max_header_len argument; shouldn't we rather put
> the default value to 0 and pass the argument to
> email.Generator.Generator.__init__? This would avoid overwriting
> _write_headers.

A matter of style.
The change in the new patch now has a children_maxheaderlen in third
position of the __init__ arguments, which is different to what the
Generator class gave in the documentation.
Again: No practical consequences.
In my patch it was quite easy to see that no folding will happen anymore.

Bottom line: no real differences, except minor stylistic ones.