[e2e] Number of persistent connections per HTTP server?

[Scott W Brim]

On Thu, Oct 10, 2002 10:10:56AM -0500, Spencer Dawkins allegedly wrote:
> A couple of weeks ago, we were looking at the Mozilla source code, and tripped over http://lxr.mozilla.org/mozilla1.0/source/netwerk/protocol/http/src/nsHttpHandler.cpp, which (today - it does change!) includes
> 
> 114     , mMaxConnections(24)
> 115     , mMaxConnectionsPerServer(8)
> 116     , mMaxPersistentConnectionsPerServer(2)
> 117     , mMaxPersistentConnectionsPerProxy(4)
> 
> and (later in the same file)
> 
> 939     PRUint8 maxPersistentConnections =
> 940         ci->UsingHttpProxy() ? mMaxPersistentConnectionsPerProxy
> 941                              : mMaxPersistentConnectionsPerServer;
> 
> So it looked (from 50,000 feet) like Mozilla 1.0 does make a distinction between number of persistent connections to a server and number of persistent connections to a proxy. No, I haven't been through the code enough to know how far the implementation goes with this thought - but it seems to be part of the plan.
> 
> Mozilla seems to be implementing features like HTTP request pipelining, so this didn't surprise us a lot - we're just curious what science is behind the distinction between "number of connections per server" and "number of connections per proxy".
> 
> Any pointers?
> 
> Spencer

Ruminations (perhaps they will inspire someone to come forward with the
real answer): IIRC the max connections parameter went into Netscape
because of issues with Windows 95 and multiple connections being
inefficient, but I don't remember how.  Proxies are certainly closer
than most servers, and probably a lot closer, so I would guess the
problem they were compensating for had to do with round trip delay.

swb