Re: Holding connections open: an immodest proposal

Marc VanHeyningen (mvanheyn@cs.indiana.edu)
Thu, 15 Sep 1994 07:40:22 -0500


> In article <866E@cernvm.cern.ch> you write:
> |>I'd not agree with this; the client, not the server, knows what the
> |>client wants. It's also very hard to make this backward-compatible,
> |>since most existing clients can't handle multipart messages (can't
> |>even hand them off to an external viewer.)
>
> No the server knows as well because the client sends an accept field.

Um, I must not be understanding what you mean.

When retrieving an HTML document, there is nothing about the
Accept: field which indicates whether the client is capable of, or
interested in, fetching and displaying inlined images (or, more generally,
any other items which are logically part of that same document.)

> Of course if the client sends Accept: */* with no quality field and does
> not know what to do with it then the system will break.

Some might say such clients Deserve To Lose. :-)

> I had an idea to deal with this, ammending the content type so that
> a */* match without quality field defaulted to q=0.01 and have a cutoff
> for sending multipart if the q factor was not > 0.1
>
> We can call it the M-Kludge.
>
> (because its needed for MIME - what did you think it stood for?)

Heh heh... of course, it's not needed for MIME, which is in the spec already,
but for broken clients. Hence the other meaning.

I'm fairly against adding kludges to the protocol in which we assume that
the client is lying in the Accept: fields. Fix the damn clients.

Marc VanHeyningen <http://www.cs.indiana.edu/hyplan/mvanheyn.html>