Re: C++ PATCH: Tree dumper

[Jason Molenda <jason-gcclist at molenda dot com>]

On Sun, Dec 02, 2001 at 07:41:47PM -0200, Alexandre Oliva wrote:
> On Nov 21, 2001, Jason Molenda <jason-gcclist@molenda.com> wrote:
> 
> > Besides which, our wonderful web archives (All Hail Mhonarc) will
> > decode and save aside any attachment with a disposition of attachment
> > into a text file.
> 
> Would it be possible for it to save aside inline parts of type
> text/x-patch, without preventing them from being displayed inline?  Or
> having all MIME parts saved separately, for that matter?

Unfortunately, this is entirely to the point of the disposition
specification -- items marked as disposition inline are intended
to be viewed as the main part of the message.

But to answer your question, no, you have two choices with mhonarc
- you can either force all inline parts to separate files, or you
can have them displayed inline.  If you select the former, the
message views are empty; they have a link to the main text's message.
(for those not familiar, the normal 'default' thing you see when
you look at a message is the 'inline' part.  This part I'm writing
right now is the inline part for this message).

But I believe I mentioned that raw messages are automatically
availble now in all the web archives as of four weeks ago, so unless
the message has been quoted in some fashion (base64, quoted-printable,
maybe format=flowed), you can get the raw message and run that
through patch without any problems.


> > For instance, Andreas Schwab sends e-mails with a charset of
> > iso-8859-15 and his Content-Transfer-Encoding is quoted-printable.
> 
> It should be possible to encode iso-8859-15 as 8bit.

Exactly, which is why his MUA is sending the messages as quoted-printable
- there may be some 8-bit chars in there.

> In any case, GCC sources are pure ASCII, so it should be possible to
> encode at least the patches as 7bit.

Unfortunately, I don't think many MUAs work that way.  You say "I will
be sending messages in language FOO", and the MUA says, "Hey, FOO has
high bit characters in it - I'll quote the contents of that message".

> If they don't know how to configure their MUA so as to post MIME
> properly, better have them to send patches as plain text.  

The whole point of MIME is to hide all of these silly details from
users entirely.  Who wants to know what their encoding of their
language is, or how it needs to be quoted to be sent safely across
the net?  For ten years MUAs have been rushing to hiding all of
this from you -- you simply shouldn't have to care.

If there are gcc maintainers who are using some MIME unaware MUA
(does anything more sophisticated than /usr/bin/Mail _not_ understand
MIME these days?) and people need those maintainers to review their
patches, then fine, tell people that they should send their patches
in with /usr/bin/Mail or suffer possible patch delays if their
mailer does any fancy MIME stuff.

Personally, I would be documenting this on the other side.  I'd
push for a statement that gcc maintainers should be prepared to
deal with MIME encoded patches.  MIME is not a radical new thing,
it is not some controversial thing, it is a message encoing scheme
designed to be hidden from people.   You're (obviously) free to
shift the burden of effort to all the people who want to send in
patches, but you're fighting a decade old movement, and expecting
everyone to know the intimate details of what their MUA might be
doing behind their backs -- I think it's a bit silly.

Jason