MIME boundaries and crypto

MIME boundaries and crypto

[Lars Magne Ingebrigtsen]

If seems that if you have a multipart message (made by Gnus with the
normal "==-=-=" boundaries, and you then forward this as a signed
message, the singing function will also use the "==-=-=" string as the
boundary, thereby rendering the forwarded message invalid.

Before I delve into the crypto signing code, does this make sense?
Does the signing functionality not ensure that the separator it's
going to use isn't already used in the body of what it's going to
sign? 

-- 
(domestic pets only, the antidote for overdose, milk.)
   larsi@gnus.org * Lars Magne Ingebrigtsen

Re: MIME boundaries and crypto

[Simon Josefsson]

On Sun, 20 Jan 2002, Lars Magne Ingebrigtsen wrote:

> If seems that if you have a multipart message (made by Gnus with the
> normal "==-=-=" boundaries, and you then forward this as a signed
> message, the singing function will also use the "==-=-=" string as the
> boundary, thereby rendering the forwarded message invalid.
>
> Before I delve into the crypto signing code, does this make sense?
> Does the signing functionality not ensure that the separator it's
> going to use isn't already used in the body of what it's going to
> sign? 

It looks like a bug.  If possible, it would be nice if Gnus _would_ keep
the MIME delimiters used within signed parts the same, and change those
outside signed parts, because the signature is calculated over the MIME
signature as well (sic!).

Another idea would be to default to forwarding signed message as <#part 
type=message/rfc822 raw=t> instead, which would preserve the original MIME 
delimiters as well.

Re: MIME boundaries and crypto

[Lars Magne Ingebrigtsen]

Simon Josefsson <jas@extundo.com> writes:

> It looks like a bug.  If possible, it would be nice if Gnus _would_ keep
> the MIME delimiters used within signed parts the same, and change those
> outside signed parts, because the signature is calculated over the MIME
> signature as well (sic!).

Yes.  The other functions that create multiparts use
`mml-compute-boundary' to compute the boundary, and it examines the
contents to make sure there are no clashes.  I'm not familiar with the
crypto code, so could somebody else fix this?

-- 
(domestic pets only, the antidote for overdose, milk.)
   larsi@gnus.org * Lars Magne Ingebrigtsen

Re: MIME boundaries and crypto

[Florian Weimer]

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

> Yes.  The other functions that create multiparts use
> `mml-compute-boundary' to compute the boundary, and it examines the
> contents to make sure there are no clashes.  I'm not familiar with the
> crypto code, so could somebody else fix this?

Is it possible to transform the MML and use the hook
'mml-generate-multipart-alist'?  I verified that it works correctly a
long time ago. ;-)

Re: MIME boundaries and crypto

[Lars Magne Ingebrigtsen]

Florian Weimer <fw@deneb.enyo.de> writes:

>> Yes.  The other functions that create multiparts use
>> `mml-compute-boundary' to compute the boundary, and it examines the
>> contents to make sure there are no clashes.  I'm not familiar with the
>> crypto code, so could somebody else fix this?
>
> Is it possible to transform the MML and use the hook
> 'mml-generate-multipart-alist'?  I verified that it works correctly a
> long time ago. ;-)

I'm not sure what you're asking.  When generating the MIME, the
boundaries have to be computed correctly.  `mml-compute-boundary' does
that. 

-- 
(domestic pets only, the antidote for overdose, milk.)
   larsi@gnus.org * Lars Magne Ingebrigtsen