Re: Two ways of implementing CID: and MID: schemes

[Ed Levinson <elevinso@xxxxxxxxxxxx>]

Harald,

What does "UA: Passes the whole multipart/related to the RA" mean?  Two
interpretations:

   1.  Pass the Mul/Rel as a single entity and let the RA parse it.

   2.  Have the UA parse and store the Mul/Rel entities and pass
       their names and headers to the RA.

Did you mean the latter?  It make more sense, allowing the
UA to remove Content-Transfer-Encodings and resolve Mul/Alt and
Msg/Ext-bod.

Best.../Ed

Thanks for the UA->RA->Browser acknoledgement.

On Tue, 21 Nov 1995 10:55:04 +0100 Harald.T.Alvestrand%uninett.no@Princeton.EDU wrote:
> (This is a resend due to a LISTSERV problem.
> It was originally CCed to the URI list, but please reply only to
> ietf-822)
> 
> Going over the recent debate, I thought I'd show two ways of implementing
> CID: and MID: URLs without modifying browser code.
> 
> Both follow the "UA->RA->Viewer" model that I think Ed Levinson
> used first.
> 
> 1) MODIFY THE HTML
> 
> UA: Passes the whole multipart/related to the RA
> RA: Stores all parts in separate files, replaces all occurences of
>     the Cid: and Mid: URL with file: URLs.
> Browser: does its thing.
> 
> 2) PROXY THE COMMANDS
> 
> This depends critically on the theory that browser implementors
> will have an easy way of using the xyzzy_proxy environment variable
> when an xyzzy: URL is encountered.
> 
> UA: Passes the whole multipart/related to the RA
> RA: Indexes the body (maybe saving it to file, maybe not), starts
>     listening on port 12345 (random), and sets mid_proxy and cid_proxy
>     to localhost:12345 before invoking the browser
> Browser: Does its thing.
> 
> ...