Competetion To Google?
Posted 30 September 2004 - 10:05 AM
Posted 30 September 2004 - 12:27 PM
Vava it is a "Competition"
Edited by ArmenSarg, 30 September 2004 - 12:28 PM.
Posted 30 September 2004 - 12:44 PM
Microsoft just started a big arse research group on search engine and it looks like they are trying to position themselves to take over this area as well big time. I know a couple of guys that just got hired there with recent PhDs.
The problem with current search engines (ala google) is that it takes some time for them to crawl the web (updates to a site might take weeks to show up) ... plus a lot of the web now is not "static" pages ... they are dynamically generated content. So the "search" arena is very very hot in terms of research both in academia and industry.
The other thing is that not too many users are willing to wait more than a second or so for their search query answer There are also some things in google's page rank algorithm that can be improved upon so I'm sure we will see some competition for google in terms of technology. The problem is going to be muscle and Microsoft definitely has that.
Then again, we are still stuck using 1970's technology for our internet ala TCP/IP. Why is it still around? Cause it's too expensive and complicted to make the switch to the much better things that have been developed since then.
Edited by Seapahn, 30 September 2004 - 12:46 PM.
Posted 30 September 2004 - 01:05 PM
Sip, what are the main features of TCP/IP and how are the new ones different from that?
Posted 30 September 2004 - 02:03 PM
Well, the basic concept of TCP/IP is sound ... it sends data in form of small packets from source to destination while not everything follows the same path. So when posting this message on the forum, part of the post may have gone through china to end up at the server while other parts may have come more directly.
The main problem with IP (v4) that is used almost everywhere is the IP address space limitation. The IP address that is a unique address assigned to every device connected to the Net is in the form of 255.255.255.255 (4 pairs of 8 bit numbers that can range from 0 to 255). Some are already reserved for specific tasks but there are only 4 billion total unique addresses possible (in reality much less) which as you can imagine is becoming very very limited.
In addition to that, there were some that were predicting the net come to a screaching halt in the mid 90s. The way TCP works is that if there is congestion in terms of packet delivery on the communication lines, it backs off for exponentially longer and longer periods of time until the congestion is relieved. Of course things are a lot more complicated and is hard to explain it in a short post but just trying to give you a flavor of "why" ... when it was invented, it was used to connect a handful of machines ... maybe hundreds or thousands. Now we don't even know how many there are!!!
Ok and now for the most amazing part ... why did TCP/IP become the defacto standard? Because it was available for free as part of the BSD Unix package long ago. Basically some poor and hungry Berkeley grad students stayed up long nights to write the code to do their masters and phd dissertations. If you look at some of the stuff they did is pretty funny! Like to do the "Internet checksum" as it is now known, i.e. to detect whether an arriving packet has bit errors, they just sum up all the bytes and do a simple modulo calculation. Well, it's easy to see that if you sum everything up, if there was a bit flip from 1 to 0 and another from 0 to 1, it'll all cancel out and you won't detect the error. Stuff like this that may have seemed like a great idea at 2AM in the morning before a deadline ... but after a couple of decades of intense research by MANY many bright minds you can imagine how they can seem quite childish.
Well, essentially the entire protocol stack that we use to communicate these days has sorta evolved on its own in slow bits and pieces and different standardizations and many different organizations having their hands in it ... as well as simply market forces and having to keep backwards compatibility in mind. There are many redundant things and there are things like security which is VERY important but were only added later so they are not really integrated into where they maybe should be.
Add on top of that things like "multicast" where one server sends the same thing to many users (as you can imagine that happens a lot) ... the original TCP/IP, wasn't designed for that so the server would have to send a different message (albit the same data maybe) to every single user that requested it. But with multicasting, that can be done much more efficiently (like for broadcasting some video to a bunch of subscribers on the web).
Posted 30 September 2004 - 02:25 PM
Posted 30 September 2004 - 03:26 PM
Posted 30 September 2004 - 10:20 PM
Multiverse would be a good name.
Posted 30 September 2004 - 10:23 PM
Well, you know what to call it if it fails - Clutzy.
Posted 01 October 2004 - 08:38 AM
I keep reading "Crusty"... I agree with you nairi, that name is just not going to make it.
0 user(s) are reading this topic
0 members, 0 guests, 0 anonymous users