- From: Henrik Frystyk Nielsen <frystyk@microsoft.com>
- Date: Thu, 25 Jan 2001 14:04:44 -0800
- To: "Mark Baker - Ottawa Consumer and Embedded Div." <Mark.A.Baker@canada.sun.com>, <xml-dist-app@w3.org>
- Cc: "Frank DeRose" <frankd@tibco.com>, <rayw@netscape.com>
Indentation: ">>>" = Henrik ">>" = Mark ">" = Henrik "" = Mark *** >> [XML has its own problems as an envelope] > >XML is really good at nesting within the same XML document but no argument >that nesting of documents is less so. However, I would note that although >MIME can contain nested MIME messages then it is not possible to do so in >a way that allows the semantics to be nested as in XML and nested MIME >multipart plainly is not simple. This is actually the reason why the SOAP >attachment spec doesn't go any further but more about this later in this >mail... > >A more fundamental problem that I have eluded to and several have asked >what I mean by is the problem of referring to MIME multipart entities. >Let me illustrate it with an example: > > I send you an insurance claim with two photos of my broken car. > I include the photos in the MIME multipart message and refer > to them using "mid:" URIs. > >However, "mid:" URIs are generally only resolvable within the context of that >MIME multipart message. That is, the insurance company cannot send me back a >message saying photo A was good but photo B didn't show any marks unless it >actually sends me back the two photos. But even then I have no way to see >that those two photos actually are the same as the ones I send in the >first place. > >Why? Because MIME message id's are not intended to be "resolvable" - they >are only intended as identifiers. > >Now, you can say that of course I should keep track of all my message >id's that I have generated over time so that I know what you are >referring to - somewhat like an NNTP server but that suddenly makes it >much more complicated to make an implementation. > >What would be just as reasonable would be to have the scenario play out >this way: > > I send you an insurance claim with references to two photos of > my broken car. I include the references to the photos using > something like "http://my.com/brokencar..." URIs. > >Of all things that HTTP is being used for this is actually a good one :) >HTTP is really good at GET! You're preaching to the choir! 8-) But I don't see why using MIME necessarily means using a mid: schemes to identify parts. Why not use the true URLs and Content-location rather than locally-scoped names? >>> That is certainly one reason but even more importantly, caching, >> >> If the Content-location header were used properly on the parts, is >> caching still a problem? i.e. is it the mechanism that's at fault, >> common use, or existing implementations? > >Well, as soon as you actually include the document in the MIME multipart >then caching is immaterial - the whole purpose of caching is that you >*don't* have to send the data. Hehe, obviously. 8-) I meant that a part could be cached with its Content-location so that if later some other non-multipart document arrived with a reference to that document, that it could be retrieved from cache. >But you touch on one of the main issues - it is exactly correct that one >can use Content-Location and this brings me to the crux of my point - >the fundamental concept that we absolutely have to have is links - >without it we can't refer to anything outside the XML envelope. > >Once we have links, they can of course refer to anything and has been >mentioned MIME multipart bodies is one such example--you mention >carrying SOAP data within a MIME multipart which is definitely a >possibility. However, they can refer to http://my.com/foo and >mailto:frystyk@my.com as well. > >The important thing to realize about having links is that at this point, >using MIME multipart becomes an optimization choice that applies to >certain scenarios but not all. > >This is exactly why SOAP made the decision to provide links but keep the >door open for what these links can point to AND to keep the door open >for providing multiple underlying protocol bindings that may support >different types of links. But to my knowledge, nothing in MIME prevents us from using URLs and Content-location to solve this problem. >> I think I'd need some examples to convince me. > >Thank you! Thanks for that response. We're getting somewhere now. MB -- Mark Baker, Sun Microsystems Ottawa, Ontario, CANADA
Received on Thursday, 25 January 2001 17:05:22 UTC