httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Roy T. Fielding" <field...@avron.ICS.UCI.EDU>
Subject Re: more Keep-Alive - possible problem
Date Wed, 07 Feb 1996 21:37:12 GMT
> This is what really annoys me about Keep-Alive, Host etc al;
> there is no (provisional) standard these for features, and if there were
> it would be called HTTP/1.1; yet browsers and servers are implementing
> their understanding of these features without changing the protocol version
> number.

Yep, I could not convince people (particularly Spyglass) that the idea
was to make an *experimental* implementation and not distribute it
to the general public.  Instead, their marketing turned it into a feature
because Netscape was out of the loop (i.e., they were not involved in the
IETF at the time Keep-Alive was worked-out).  That also destroyed
interoperability because now Keep-Alive cannot be trusted if it passes
through a proxy.  So, we are stuck with inventing a new directive for
the Connection header for a proxy keep-alive.

> Netscape's behaviour is correct (even if it does highlight the problems
> of Keep-Alive in 1.0); it should ignore the Content-Length header
> if it conflicts with the amount of data sent by the server, as that
> header cannot be considered reliable in HTTP/1.0.

No, it is incorrect -- one of the requirements placed on NCSA and Spyglass's
implementations is that the Connection: Keep-Alive also guarantees a
correct content length.  If more data arrives, it should not be displayed
because the user needs to know that something is wrong with the server.


View raw message