Re: uh oh -- halp!

Marc Andreessen (marca@ncsa.uiuc.edu)
Thu, 9 Sep 93 00:15:07 -0500


Matt Heffron writes:
> >> I have an option. Don't mess with the CR LF stuff, just add a way
> >> to indentify the protocol. If HTTP servers identified themselves at
> >> startup by saying "welcome: HTTP/1.0", or something like that, then we
> >> would know exactly what to send and expect. This would solve alot
> >> of current problems and some BIG future problems as well.
> >>
> >> :lou
> >
> >But won't this require 2 round trips for every document. First ask
> >the server its HTTP version, then send the request. Massive slowdown
> >it seems to me.
> >
> > Eric
>
> Worse than that, clients that CAN take HTTP/1.0 will wait for a
> REAL LONG TIME for HTTP0 servers to identify themselves. ;-)

The proposed solution (most nested paragraph above) is indeed not
tenable; the mutual client/server identification that's a part of
HTTP/1.0 and (ya gotta hope) later revisions of HTTP covers this
aspect of the situation anyway, while only requiring one round trip.

Marc