Re: Persistent Connection Question

Jeffrey Mogul wrote:
>    3) Some transport stacks have a more difficult time detecting a
>    closed remote connection than a reset one (The reset causes all
>    calls to report an error while a close causes many to continue to
>    work).
>Since HTTP/1.0 uses close-of-connection as the end-of-entity-body
>indicator, how could such stacks support HTTP/1.0?  This seems
>so fundamentally broken that I find it hard to believe we should
>warp the protocol to deal with it.
Jeff, I think you misunderstand what I was saying. I am not suggesting a
reset after the entity-body. I am suggesting allowing a reset _from the
client_ rather than the client sending a subsequent request.

I agree that the server must use a close rather than a reset, unless it is
responding to a network failure, etc.

Alex Hopmann
ResNova Software, Inc.

Received on Tuesday, 26 March 1996 21:57:27 UTC