Speed up your browsing
- Thread starter DocBot
- Start date
...although you can set the amount of tons, here. And I still don't know what's so bad with it. Of course that's mainly due to ignorance, but I still get this feeling of "Nooooo! The bandwidth! Preserve the precious bandwidth! It is important to have, in case of, er, not enough spare bandwidth!"
Actually, it slightly reduces the bandwidth used (as it gets rid of some packet overhead.) And when used responsibly, it's mostly ok. I'd keep it around the default 4 or 5.
The thing is, it doesn't make a whole lot of sense to use on low-latency connections (cable, DSL.) It's really more for high latency connections (satellite and the like) where the wait time for each response from the server is significant.
Essentially, you're causing the server (and your connection) to have to deal with one big chunk rather than many small ones. The internet is a series of tubes, remember, and it would prefer to deal with small requests where possible.
This all based on my probably-not-comprehensive knowledge. Just remember that there is a reason it is disabled by default
The thing is, it doesn't make a whole lot of sense to use on low-latency connections (cable, DSL.) It's really more for high latency connections (satellite and the like) where the wait time for each response from the server is significant.
Essentially, you're causing the server (and your connection) to have to deal with one big chunk rather than many small ones. The internet is a series of tubes, remember, and it would prefer to deal with small requests where possible.
This all based on my probably-not-comprehensive knowledge. Just remember that there is a reason it is disabled by default
Wol
In Cryo Sleep
Am intrigued to know why this "overstresses" servers. Its not like it loads up the page several times over.
edit: reading the wrong variable and went down the wrong road after hearing people talking about concurrent connections
Pipelining, just sticks several requests down one connection, rather than multiple connections i thought? Saves the server having to open and close the tcp connections, which is usually the slow bit. Just some servers get a bit confused that after theyve worked with one request, they recieve another, and then go "meh, ill just timeout".
And id disagree with the thing about the internet preferring to deal with small packets. small packets mean that theres more headers, opening and closing of connections going on. Id prefer to , if i had a connection, keep it. Imagine coding a small server. You could either have something which opens a connection, then loops sorting out the requests through this connection, then closing it. Or, you could have a loop, which opens and closes the connections for each request. The first one would be better in my experience, as the program can spend more cycles sending data, rather than managing sockets.
in the same way that in online games, generally one socket is opened, and all the data is sent through that when required, rather than a socket being opened and closed each time you want to send something?
The reason it is disabled by default would seem to be that some webservers are a bit silly, and dont support multiple requests through one connection, and therefore get confused, and die.
this is based on my knowledge of adminning webservers, and from doing a fair bit of programming using sockets and network connections
edit: reading the wrong variable and went down the wrong road after hearing people talking about concurrent connections
Pipelining, just sticks several requests down one connection, rather than multiple connections i thought? Saves the server having to open and close the tcp connections, which is usually the slow bit. Just some servers get a bit confused that after theyve worked with one request, they recieve another, and then go "meh, ill just timeout".
And id disagree with the thing about the internet preferring to deal with small packets. small packets mean that theres more headers, opening and closing of connections going on. Id prefer to , if i had a connection, keep it. Imagine coding a small server. You could either have something which opens a connection, then loops sorting out the requests through this connection, then closing it. Or, you could have a loop, which opens and closes the connections for each request. The first one would be better in my experience, as the program can spend more cycles sending data, rather than managing sockets.
in the same way that in online games, generally one socket is opened, and all the data is sent through that when required, rather than a socket being opened and closed each time you want to send something?
The reason it is disabled by default would seem to be that some webservers are a bit silly, and dont support multiple requests through one connection, and therefore get confused, and die.
this is based on my knowledge of adminning webservers, and from doing a fair bit of programming using sockets and network connections
Disclaimer: this entire post is based on the vauge memory of something i wrote about 3 years ago so may be entirely wrong
i believe it's mainly to do with the way sockets are handled ( and also the way windows servers don't handle them)
basically for every connection to the server it uses one of it's sockets, however when you close a connection that socket still has some of a set length of time till it times out (iirc its 5 minutes) so if you have 10 connections that reduces the number of connections that server can make by 10 now there are 65534 sockets available which means if every person who visited a site used this method the server could only handle ~6000 people over the 5 minute period hence why you run into problems with the server appearing to be unavailable especially when you have things like php and ajax which cause extra connections to a server anyway
i believe it's mainly to do with the way sockets are handled ( and also the way windows servers don't handle them)
basically for every connection to the server it uses one of it's sockets, however when you close a connection that socket still has some of a set length of time till it times out (iirc its 5 minutes) so if you have 10 connections that reduces the number of connections that server can make by 10 now there are 65534 sockets available which means if every person who visited a site used this method the server could only handle ~6000 people over the 5 minute period hence why you run into problems with the server appearing to be unavailable especially when you have things like php and ajax which cause extra connections to a server anyway