Status of MHTML implementation in the spring of 2000

Two students at Stockholm University have as their master's thesis made an evaluation of the support for the MHTML standard by some mailers. Their full report is available in plain text and Word format. Below is a very short summary.

Tested systems

Fuller name

Test platform

Version

Short name

Microsoft Outlook Express

Windows NT

5.0

Outl

Netscape Messenger

Windows NT

4.7

Mess

Qualcomm Eudora Pro

Windows NT

4.2.2

PC Eudo

Qualcomm Eudora Pro

Macintosh (not fully tested)

4.2.1

Mac Eudo

Pine

Compaq Alpha Unix

4.20

Pine

Juno

Windows NT & Explorer 5

4.05

Juno

Microsoft Hotmail

Explorer 5 from Windows NT

April 2000

Hotm

Yahoo! Mail

Explorer 5 from Windows NT

April 2000

Yaho

KOM 2000

Explorer 5 from Windows NT

3.0

KOM

MHTML facilities support

Facility

Can produce

Can receive

Multipart/related and Message-ID links to images

   
  

Without multipart/alternative

Outl, Mess, PC Eudo, Hotm

All

Multipart/alternative inside multipart/related

Outl, PC Eudo, Hotm, Yaho

Mess, KOM, Mac Eudora (Not tested for Outl, PC Eudora, Pine, Juno Yaho)

Multipart/alternative outside multipart/related

Mess

Mess, Hotm, KOM, Mac Eudora (Not tested for Outl, PC Eudora, Pine, Juno Yaho)

Multipart/related and Content-Location links

Outl

 
  

With absolute URLs

 

All except Hotm

With relative URLs, no base

 

Outl, Pine, Hotm, Juno

Relative URLs, base in Content-Location in a surrounding heading

 

Outl, Pine, KOM, Juno

Relative URLs, base in Content-Location in its own heading

 

Outl, Mess, Pine, Juno

Content-Location to be recursively applied to inner parts

 

Outl, Pine, Juno

Resolution through HTML <BASE> element

 

All except Hotm

Multipart/alternative inside multipart/related

Outl, PC Eudo

Outl, Pine, Hotm, Juno

Multipart/alternative outside multipart/related

Mess, Hotm

Outl, Pine, Juno

Unfolding of long URIs in MIME headers

 

Outl, Pine, Hotm, KOM, Juno

Decoding of encoded characters in URIs in MIME headers

 

Hotm, Mess, Pine, Hotm, KOM, Juno

Progressing MHTML to draft standard?

Issues to discuss on whether to progress this standard from proposed to draft status:
  1. Only one single of the tested implementations (Outlook Express) produces messages using Content-Location between body parts. This feature was added to the standard, in order to allow existing web pages to be sent through e-mail without modification of the HTML text. The reason why this is only supported by one implementations is that for some time, one major implementation (Netscape Messenger) did not support receipt of this format. Because of this, other implementations did not want to produce it, because their messages would then not be receivable with Netscape Messenger. However, Netscape Messenger now supports receipts of Content-Location. In fact almost all implementations now support receipt of this format. The reason for this is probably that since one major implementation (Outlook Express) sends in this format, all implementations want to support receipt of this format. Should we keep this in the standard, even though only one implementation produces this format? Should we encourage some more implementation to support this format? For example, a small not-general-purpose program just for mailing web pages and nothing else?

  2. Half of the implementations do not support the variant of Content-Location containing relative URLs. The variant supported by almost all implementations for receipt is Content-Location containing absolute URLs. Should we change the standard to only include this variant?

  3. Many major implementations support combination of multipart/related with multipart/alternative. This combination, however, is used in two different ways. Netscape Messenger sends messages with multipart/related inside multipart/alternative, all other implementations send with multipart/related outside multipart/alternative. For receipt, this combination is not well supported except with Content-ID relations between body parts, but done that way, several implementations can receive both variants. Rather strange was that Hotmail could receive only the test message with multipart/alternative outside, but could produce only messages with multipart/related outside! Should the standard say something about this combination? To get better interoperability, it might be specifically required that a conforming implementation must be able to receive both variants of the use of multipart/alternative?

  4. The testers did not succeed in producing any messages with the rules for encoding of special characters and long rows using the tested mail clients. Should this facility be removed from the standard?

By Jacob Palme, First version June 2000