Re[2]: Re: Firefox users - please note....

["Pete Shew" <ukha_d@xxxxxxxxxxxxxxxxxx>]

I seem to recall that IE has a concurrent figure of four sessions by
default but limited to two per server. I'm surprised Firefox was set to a
single session by default.

Pete

*********** REPLY SEPARATOR  ***********

On 11/01/2005 at 11:32 James Fidell wrote:

>Quoting mark_harrison_uk2 (mph@xxxxxxx):
>>
>> --- In ukha_d@xxxxxxx, "J."
<chemically_enhanced@h...> wrote:
>> >
>> > Unless it happens to be your server that people are
accessing. most
>> > servers ar throttled to handle a mximum number of request at
any one
>> > time,  If you put your settings at 30 - 40 and everyne else
does it
>> > acts as a denial of service to other users as the server will
just
>> > return a 503 http response. play nicely!
>> > #
>> > </rant mode off>
>> >
>> > J.
>>
>> Have you seen this happen? The discussion lists have been full of
>> people warning that "this could happen", but I've not
yet seen anyone
>> actually demonstrate that it HAS happened.
>
>My understanding of what happens is as follows:
>
>Normally when you request a web page, your browser opens a single
>connection to the webserver and sequentially requests all the elements
>of the page from that single connection.
>
>What the "pipelining" stuff appears to do is to tell Firefox
to open
>multiple concurrent connections to the webserver at the same time and
>"multiplex" all its requests down those connections.
>
>From the webserver's point of view this means that it will be trying to
>serve lots of different files to you at once, as opposed to one after
>the other.  Webservers however usually have a fixed limit on the number
>of concurrent connections they'll allow in total, so if enough people
>are opening lots of connections to the same server at the same time,
>the server will start returning a "too busy" message when
there are
>fewer users than if everyone just opened one connection at a time.
>
>One might think that this shouldn't actually be a problem, because
>although you're opening multiple connections, you're requesting fewer
>files from each one and therefore not using the connection for as long.
>However, it's usual once a client has connected to a webserver to keep
>the connection open for a while in case it needs to be used again,
>because doing so requires less resources than closing the connection
and
>re-opening it later.  So, even after you've grabbed the web page you
>may still be using resources on the webserver.
>
>This may result in an effective denial of service for users of the
>website, depending on how busy it is. If that does turn out to be
>the case, I'd expect the server administrators might reconfigure the
>server to restrict the number of concurrent connections from any given
>client and/or reduce the time that the connection is kept alive. I
>shouldn't be surprised if that resulted in people who are using the
>pipelining finding that their browser performance has returned to
>it's pre-pipelining levels and those that don't use it finding that
>theirs has reduced further. If it becomes a serious problem I guess the
>developers will just remove the settings from future releases of the
>browser.
>
>All this is theory though.  What's really required is some empirical
>evidence, which seems unlikely to be forthcoming given that it'll be
>hard work to collect.
>
>James
>
>
>
>
>Yahoo! Groups Links
>
>
>