Yes! MIME readers must not need to understand html. A model that
helped me separate the issues (wrt SGML) has three elements, not two.
+------------+ +------------+ +------------+
| MIME | | Receiving | | Web |
-->| User Agent |----->| Agent |----->| Browser |
^ +------------+ ^ +------------+ ^ +------------+
| | |
|<-messages |<-data files |<--{<url,file>}
|<--headers
Using this model the MIME UA (MUA) just deals with the body parts
and passes the names of the data files and files containing the
headers to a Receiving Agent (RA, an impdedance matcher between
the MUA and the browser). The RA may require knowledge of the
browser and HTML. In my view it examines the Content-Disposition
(cDisp) headers and moves the corresponding data to the browser's cache.
After doing its job the RA invokes the browser. The RA takes care of
mapping URLs into local file names, something that should not concern
the sender.
The above processing implies two assumptions, the browser has a cache
and the cDisp header contains absolute URLs. Additionally, on
the sender's side a corresponding Sending Agent, which must
understand HTML, provides the absolute URLs.
In this model the RA is the helper app. At some point it may be
incorporated into the browser but the model supports existing browsers.
Best.../Ed
On Thu, 02 Nov 1995 19:01:49 EST Keith Moore wrote:
[Editing the text/html body part]
is what I'd like to avoid. I don't think the MIME reader should
have to know about text/html (or any other body part) so it can convert
its internal references to another format. I'd far rather the helper
apps have some interface to map content-ids to filenames (which the
MIME reader has conveniently stored), but this requires making the
relevant helper apps smarter, which is difficult. Using content-disposition
filenames seems like the easiest hack which doesn't require substantial
changes to the helper apps.
However, it doesn't work to put arbitrary URLs in the Content-Disposition
field, because most file systems won't be able to store files using
a URL as a file name. The filenames used for this purpose need to be
very portable, say xxxxxxxx.yyy format using upper case letters or digits,
and maybe with certain names disallowed.
Keith