Re: SPEC: DNS conventions & "naming" of VRML sites

Anthropohedron ([email protected])
Wed, 14 Jun 1995 14:36:17 -0400 (EDT)


} Re, the "not tied to content types"? How is a Web document
} able to indicate that it is composed of multiple types
} and which version or flavor is to be used? For example,
} if I have a document with links to HTML, VRML, any SGML,
} JAVA, etc.? Is the file extension the only indicator
} and if so, how is that able to indicate the flavor?

There is a MIME type, multipart. The subtype multipart/mixed is commonly
used in mail to send text with graphics or other non-text. Some mailers
even separate the .signature by making the li> multipart. Netscape, by
the way, has proposed and implemented multipart/x-mixed-replace for their
"server push" feature/kludge/bloat/mistake (take your pick), but does not
understand straight multipart/mixed (of course, you can set something up in
your mailcap, like piping it through metamail, but that is a poor
solution).

I think it would be profitable to start using multipart/mixed for VRML with
a text or HTML caption, and I would like to see a browser smart enough to
separate a multipart document. It might also be useful to make a new
subtype, say multipart/x-choice (the x prefix denotes an experimental
type), that would send VRML as one part and HTML as the other and the
browser would figure out if it knew how to deal with VRML, otherwise it
would read the HTML. On the other hand, the browser sends HTTP_ACCEPT along
whenever it connects to a server and it would use li> bandwidth and stick
to the lodel better if the server would check for x-world/x-vrml among the
HTTP_ACCEPT strings and decide which to send on its end.

This requires, of course, that the browser send a HTTP_ACCEPT string.
Netscape seems to send "*/*, im/gif, im/x-xbitmap, im/jpeg" as its
string, which is entirely useli> . es"sumaly everything that is defined in
whatever mailcap is used should be what is sent as the HTTP_ACCEPT.

Finally, I think the whole thing will be simplified by merging some HTML
browser with some VRML browser. I think that inlined VRML would suit
everyone a lot better than either of these MIME-based solution . Is anyone
planning to integrate VRML into a WWW browser? Does anyone else believe, as
I do, that VRML is/should be a primary format, like GIF, JPEG, HTML, XBM,
and plaintext, not a secondary format like audio and animation?

} Len Bullard
--Greg