Re: [lamps] S/MIME fix

[Jim Schaad <[e-mail]>]

I will note that unless multipart/mixed has some special text associated with it.  Text/html is defined by RFC 1866 to only hold a single HTML document.  Thus building an HTML document from multiple parts is not according to Hoyle.

 

Jim

 

 

From: Spasm <[e-mail]> On Behalf Of Alexey Melnikov
Sent: Wednesday, May 16, 2018 7:49 AM
To: Phillip Hallam-Baker <[e-mail]>; SPASM <[e-mail]>
Subject: Re: [lamps] S/MIME fix

 

Hi Philip,

On 16/05/2018 15:28, Phillip Hallam-Baker wrote:

Looking at eFail, surely the simplest fix is to require that an HTML message body be presented in a single CMS envelope presented in a single MIME part?


I am not sure what you mean here. A CMS envelope can contain multipart/mixed within it, which is a perfectly valid use case (i.e. if one wants to send some encrypted text together with some encrypted attachments).
If you are talking about preventing the following construct:


content-type: multipart/mixed; boundary=.f8231d7f-681b-442c-97cc-e6c5375d059d

This is a multipart message in MIME format. 

--.f8231d7f-681b-442c-97cc-e6c5375d059d
content-type: text/html
 
....some partial HTML...
--.f8231d7f-681b-442c-97cc-e6c5375d059d
content-disposition: inline; filename=smime.p7m
Content-Transfer-Encoding: base64
content-type: application/pkcs7-mime; name=smime.p7m; smime-type=enveloped-data
 
....encrypted HTML...
--.f8231d7f-681b-442c-97cc-e6c5375d059d
content-type: text/html
 
....some partial HTML...
--.f8231d7f-681b-442c-97cc-e6c5375d059d--


i.e. a multipart/mixed that contains a mixture of text/html and application/pkcs7-mime, then I might agree with you. But this is not really an S/MIME feature, it is a generic MIME feature. So maybe this WG should write a document on best S/MIME implementation practices.




This would simplify the code substantially. While it is conceivable someone has worked out a way to make use of this mis-feature, I for one cannot imagine why Outlook, Thunderbird or the like would ever do anything of the sort.

 

 

Separately, we have interest in CAA for S/MIME. Surely we should do ACME for S/MIME as well.


Not surprisingly, I agree. See draft-ietf-acme-email-smime-02 

If we are going to do that, surely we should have a discussion of what it would take to make end to end security the default for SMTP.

 

I am not necessarily thinking of this as a LAMPS thing because we also need to get CAs, probably CABForum involved and maybe the OpenPGP folk.


Best Regards,
Alexey