Re: HTTP/1.1 Pipelining

• Messages sorted by: [ date ][ thread ][ subject ][ author ]
• Next message: M. Hedlund: "Re: draft minutes, HTTP-WG meetings April 7"
• Previous message: Phillip M Hallam-Baker: "Re: HTTP practice and Year-2000: the bad(?) news"
• In reply to: David W. Morris: "Re: HTTP/1.1 Pipelining"
• Next in thread: Jeffrey Mogul: "Re: HTTP/1.1 Pipelining"
• Reply: Jeffrey Mogul: "Re: HTTP/1.1 Pipelining"

> 
> Yes ... but there are some reasons why it may not be advisable from a
> implementation design perspective.  Consider the case where request
> below to foo.com takes forever to discover the DNS name is invalid. 
> The response from bar.com must wait and the requesting client has no
> clue why.
> 
> Dave Morris
> 

I agree that its a difficult implementation to handle, but the case you
mention isnt the one I was thinking about.  
What do we do with sticky headers?  Should each origin server get
1 copy of the sticky headers? Should the client have to specify
sticky headers for the 1st request to a new origin server in a pipeline?


-----------------------------------------------------------------------------
Josh Cohen				        Netscape Communications Corp.
Netscape Fire Department	     	       "Mighty Morphin' Proxy Ranger"
Server Engineering
josh@netscape.com                       http://home.netscape.com/people/josh/
-----------------------------------------------------------------------------

• Next message: M. Hedlund: "Re: draft minutes, HTTP-WG meetings April 7"
• Previous message: Phillip M Hallam-Baker: "Re: HTTP practice and Year-2000: the bad(?) news"
• In reply to: David W. Morris: "Re: HTTP/1.1 Pipelining"
• Next in thread: Jeffrey Mogul: "Re: HTTP/1.1 Pipelining"
• Reply: Jeffrey Mogul: "Re: HTTP/1.1 Pipelining"