W3C home > Mailing lists > Public > xml-dist-app@w3.org > January 2001

Binary attachments to XP: or unipart vs. multipart

From: Frank DeRose <frankd@tibco.com>
Date: Wed, 24 Jan 2001 11:59:20 -0800
To: <xml-dist-app@w3.org>
Message-ID: <OFELJFDBDMKCBMENENFOCEKIDCAA.frankd@tibco.com>
David,

Thanks for your clarifications on binary data. I might put things a bit
differently. In particular, I think the issue of binary data is only one of
several issues.

Some participants want a unipart approach. This approach is based on SOAP,
it seems to be the approach mandated by the XP charter (although I haven't
yet read the responses to reviews you cite below), and it is, in fact, the
approach the XP WG has taken up until this point. In the unipart approach,
the payload is a single XML document (SOAP envelope) containing a header and
body.

Other participants want a multipart approach. This approach is similar to
the more B2B-oriented protocols, like RosettaNet, BizTalk, ebXML, and SOAP
with Attachments. In the multipart approach, the payload is a MIME
multipart/related message with (possibly) multiple entities. The unipart
approach nests naturally inside the multipart approach: the SOAP envelope,
for example, could become the root entity in the MIME message and serve as
its "header" (details, of course, tbd). That is, it could point to other
entities in the MIME message. IMHO, convergence between the various
B2B-oriented protocols could be achieved on the basis of the multipart
approach.

The advantages of the multipart approach are:

1.) It can accomodate binary data without requiring that it be encoded.
2.) It can accomodate multiple XML documents.
3.) It allows standard S/MIME encryption and digital signatures techniques
to be used.
4.) It makes it easier to parse/process different parts of the payload
independently of each other through standard mechanisms.

The disadvantages of the multipart approach (quoting Henrik) are:

1.) It makes it impossible to refer to entities with reasonable URIs.
2.) It makes caching impossible.
3.) It complicates access authentication.

[Henrik said these were "a few" of the disadvantages of the multipart
approach. I wish  he would explain his meaning more fully and also list any
other disadvantages of which he is aware.]

I am not suggesting that the multipart approach should replace the unipart
approach in the XP WG. I do, however, think standardization of BOTH
approaches is desirable. The standardization of the unipart approach is what
the XP WG is currently working on and it should continue working on this
problem with all speed. But, the question of how to standardize the
multipart approach must be addressed. I don't know whether such a project
would fall within the scope of the XP WG. Perhaps, the XP WG could join
together with other bodies that are working on the various other B2B
protocols. At any rate, to judge from the messages on the mailing lists over
the last week, I would think that there are going to be a lot of vendors who
aren't satisfied until some standard for the multipart approach emerges.

Frank DeRose
TIBCO Software Inc.
3165 Porter Dr
Palo Alto, CA 94303
650-846-5570 (vox)
650-846-1267 (fax)
frankd@tibco.com
www.tibco.com
Received on Wednesday, 24 January 2001 14:57:32 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:58:58 GMT