Re: The Silent Scream: The Sad Story of Aborted Connections

Danny Padwa (padwad@psd.gs.com)
Mon, 12 Dec 1994 23:29:01 -0500 (EST)


>>>>> "Simon" == Simon E Spero <ses@tipper.oit.unc.edu> writes:

Simon> Executive summary: Preliminary results show that Netscape
Simon> aborts 8 times as many connections as NCSA Mosaic, and that
Simon> nearly a quarter of all Netscape conceptions result in
Simon> abortion.

Well.....I think someone just won the "most eye-catching Subject: line
of the year" award!

I don't think the above statistic is so shocking.......and I think
that your opinion of how bad it is depends upon your perspective.

>From a server perspective, these truncated connections are bad. From
a client perspective, they are probably good, since they mean that
(possily limited) bandwidth and rendering time is not wasted on
something I don't care about.

One of the prime advantages of NetScape as a general "Surfing" tool is
the ability to navigate off of pages before everything has loaded.
This is important both with long pages (eg the old NCSA "What's New")
where you have a hotlink at the top and don't want to wait for the
bottom to appear, or when you stumble across a page with an inlined
megabyte of worthless GIF. Should the user be held captive while the
image transfers to save the server the abort, or should the user be
able to navigate around the image, saving significant time on his/her
14.4K link?

Are we looking at some sort of "I no longer care about this data"
message from client to server that can be used to stop a pipe? This
kid of "clean" shutdown might be a little cleaner on the client side,
and certainly on the server side. If the client could (during a
transmission) send some sort of urgent data to the server telling it
to stop sending, that might solve the client's need to save bandwidth
without hitting the server too hard.

Another feature for HTTP-IV? :-)

- Danny

-------
The above does not necessarily represent the views of my employer.