Slashdot Mirror
Science Project Quadruples Surfing Speed - Reportedly
johnp. writes "A computer browser that is said to least quadruple surfing speeds on the Internet has won the top prize at an Irish exhibition for young scientists, it was announced on Saturday. Adnan Osmani, 16, a student at Saint Finian's College in Mullingar, central Ireland spent 18 months writing 780,000 lines of computer code to develop the browser. Known as "XWEBS", the system works with an ordinary Internet connection using a 56K modem on a normal telephone line.
" A number of people had submitted this over the weekend - there's absolutely no hard data that I can find to go along with this, so if you find anything more on it, plz. post below - somehow 1500 lines of code per day, "every media player" built in doesn't ring true for me.
He wants to study computer engineering in Harvard University and eventually set up his own Internet or computer company.
(For people who don't get it, Harvard's CS department, while reasonably good, is not exactly the obvious top pick among CS hotshots.)
I wonder if he will open-source the code?
it is doubtful that he will - according to the article he has applied for a patent on it.
The only way that I can see to improve the speed of browsing on the client side is to improve how fast the browser renders the page. The only other way to speed it up that I'm aware of is to change to server to also listen on UDP and have a client use UDP.
I've heard of tools in the past that claim to speed up browsing by cacheing ahead. These tools follow links on a page before you request them so that they are already in the browser's cache when you come to click on a link.
The other possibility is some heavy compression server side, but this would require a server module (e.g. mod_gzip) and this rules out any kind of built in compression in ppp, so the sppeedup would, I guess, not be as noticable as 5x.
Needless to say, I'm fairly sceptical that this is an actual speedup of browsing. If you can only fit 56Kbps down a line then you can only fit 56Kbps down a line...
It's curious that there is so few info about Adnan Osmani.
I however found out this thread in the news but, mind you, it's based on the same story...
They bet that if it's possible, he may have either implemented some quick prefetch and/or pre-formatting subroutine...
Trolling using another account since 2005.
If it does require a server side piece, it's not a web browser, per se; but as a general question, is it worthwhile to look into "compressed" web pages, e.g., foo.html.zlib?
This already exists, look for example at mod_gzip for Apache. This will compress pages before transmitting if the browser claims to support it. Mozilla does, I believe IE does too.
Sure is. So much so, that its already been done. Mozilla, for example sends a HTTP header Accept-encoding: gzip, deflate, compress;q=0.9. If the server understands that (e.g. Apache with mod_gzip), it's free to compress the data on the wire. IE (as of 5.5 anyway, don't know about 6.0) doesn't appear to send any "Accept-encoding" headers. I'd very surprised though, if this led to anything like a 400% speedup in anything but highly controlled test conditions.
I'd hazard a guess that this new browser is quietly doing some background-caching. What articles I could find about this, however, are short on detail and kinda long on BS (web browsing and watching DVD's at the same time is a revolutionary feature? riiight), so it's really difficult to tell what substance there is behind all this. Time will tell, though...
Its probably not fair to characterize Sarah Flannery's work as having had, "no solid documentation." As this page at Cryptome points out, Sarah's work did not "revolutionize cryptography" because several mathematicians -- including Sarah herself -- identified a "definitive attack" on the technique described in her winning paper (which was an application of the Cayley-Purser algorithm). Her book remains a good read, especially for young women, and I don't think anyone believes that the math in her original paper is anything less than exceptional for a 15-year-old.
The encryption story wasn't snake oil, and had very solid documentation. Sarah Flannery won Irish young scientist of the year, and subsequently the EU-wide prize, for her work. Her paper is here.
The Cayley-Purser algorithm she developed was subsequently shown to have security flaws; I don't recall if this was before or after the EU prize, but thats immaterial, the work was original and interesting, and worth a prize for a 16 year old!
She has subsequently written a book , which is a pop science introduction to crypto, and I understand from the blurb she's now studying maths at Cambridge.
-Baz
Bullshit. Get your facts straight before you malign someone. Sarah Flannery
She used Mathematica, so the Wolfram website has review of the book.
Here's a quote from Bruce Schneier in his 15 Dec 99 newsletter .
All of this was easily found with a Google search that garned 24,000 hits.
The competition is real, the prize (3000) is real, the winner is real, but I have my doubts about the project. Well, to put it differently, I think it's bullshit. Anyway, here's the news article about this. RTE is the Irish state broadcaster, BTW: http://www.rte.ie/news/2003/0110/9news/9news11a.ra m
One possible explaination for the LoC count may be that he's using Borland and trusting it's "count". At my first real job, we used Borland and I made a realtively complex program over the course of 18 months (coincidentally enough). The line count was over 1.5 million, but the reality was that it wasn't that long, Borland was counting lines processed, which included the header files, and the OWL and windows headers could add a lot to each module (of which there were over 100, since I was big on modularization).
I never really knew the true line count. I just remember the Borland one because I used to often do a global compile any time I wanted a half hour break ("Oh, the systems acting funny. Better do a global compile to make sure it's not a dependancy problem." If my boss came by and I wasn't there, he'd see the compile running on the screen).
-no broken link
The competition he won is the same one Sarah Flannery won, the ESAT young scientist competition. See:
http://www.esatys.com/
Is it possible he counted 780,000 loc because he was including libraries and component code etc. etc. The article is badly written and doesn't give a true representation of his work. He claimed on Irish TV that he had written a client-server pair. I'm still fairly suspicious myself, but it *is* possible.
this is all i could find: google groups
The article mentions that this fellow has applied for a patent last thursday. Guess what? There's no mention of the patent on the Irish Patent office website or the European patent office website or WIPO website.
not to be flippant, but since they were talking about 56k modems, which get about 4k/s average, and a typical (ide) harddrive has a (sustained) data throughput of approx 10mb/s that gives a speed-up of (10*1024*1024)/4 = 2,621,440x faster, more for SCSI, although the time taken to search the fat for the addresses of the chunks would probably bring it down a bit...
"Success is based on knowing how far to go in going too far"
Including testing, debugging and jerking off (hey, he's 16 right?)
They don't jerk off in Ireland; they have wanks.
I see a bright future for this Mr Osmani... in the internet fraud business. He's already shown his talent for overstatement and con artistry. This story would sound so believeable to someone who has no clue about how the 'net works. I doubt he has much in the way of programming skills, though.
The opinons expressed are those of the voices in the author's head and are not necessarily those of the author.
It's the avatar that gets me, though. Just can't see how (or even why) it's there.
That would be the Microsoft Agent.
Agent is simple to use. There are several dozen "agents" you can easily download that are ready to use, or you can make your own fairly easily. Here is a module I created to use the Agent in Visual Basic almost 4 years ago. Notice how easy it is to animate the Agent, and make it interactive. Once the character is loaded, you can make it do almost anything with a single line of code.
The code for the agent module can be found here.
I visited his stand at the exhibition - unfortunately, he was not there at the time - there was a note on the stand saying that he was "busy giving press interviews"!
What was displayed on the stand was very low on details as well. There was no detailed description as to how his code did what it claimed - all his paper said was that it was the "XWebs Algorithm" that did the magic! Indeed, there wasn't even a demo browser running on the stand! The only thing that I could pick up is that it seems as if he prioritises requests - though I'm not sure how the prioritisation decision is made. He also seems to make a number of simultaneous DNS requests for the one address! (gack)
However, all is not lost. He claims to have made code that generates thumbnails of web sites better than Microsoft do it (I wasn't aware that Microsoft do that, but there you go). He also has the claim of all media formats supported, as well as a built-in DVD player. I think it might possibly be an interesting product, but more from the UI experience than the speeding-up of the download of data.
well,
:-)
....
... e.g. using auto_ptr templates in C++ instead of manual exception management and manual allocation and deallocation inside of a function lets you "work" much faster and yields more maintaneable code. More readable, less to think about and faster ongoing to the next feature.
when I had "software engineering" in my computer science courses, we got this figures for LOC per say:
Application programs: 25 - 100
Service programs: 5 - 25
System programs: 1
Application programs are things like an editor (albeit some editors are rather complex), service programs are things like cc and ld or asm (albeit some of them are not "that" complex) system programs are stuff like the kernal itself or, dynamic link loaders, device drivers etc.
Well,
we all know that LOC is not a defined "value" but people working a lot with that "measure" just define it
E.g. if you work with COCOMO or with PSP(personal software process) the typical LOC is defined as a single definition, a single expression(some even say every part of an expression), an argument to a function call, every include, every define and so on:
fprintf(stderr, "this is an error number: %ld", errnum);
That would be 4 LOC, one LOC for the "statement" and 3 for the 3 arguments. Consider you can make an error/bug in every argument or 'misstype' fprintf for fscanf
LOCs do not realy get interesting in comparing hero programmers (10 to 20 times more effective) with standard programmers, but by comparing programming languages!!!
The VERY INTERESTING point about LOCs is that the noted rules of thumb above are independend from programming languages!!!
A programmer writing lets say 12 LOC per day C also writes ~ 12 LOC per day in assembler, in LISP in PERL or what ever language is appointed for the project.
So: the more expressive and the more abstract a language is the more "algorithm" or "computation" is defined in the lines of code.
In other words: 10 lines of C are far more calculation than 10 lines of assembler, while 10 lines of LISP, SQL or Prolog are even more than C.
Bottom line: the number of statements the average programmer can write depends far more on the problem domain than on the language choosen!
Well, the productivity of the so called hero programmer is in general not in lines of code, but in "abstractions" he implemetns. Or in number of features he implements. And that is often acomplished by choosing the right language constructs(not by writing more lines)
angel'o'sphere
Cost free eBook I read (by iBook/Kobo/Amazon/ObookO/Gutenberg etc.): "The Green Odyssey" by Philip Jose Farmer.
As for his claims, well, I wasn't at the show this year, so I haven't seen his entry, unfortunately. They do sound fairly unbelievable, but you have to remember that they're being filtered through journalists, most of whom are really fairly tech-ignorant.
I can say though that the Young Scientist is a major and well respected competition. The quality of the winners varies a lot from year to year, as you'd expect, but it's not run by idiots likely to be taken in by a hoax. Two yeras ago they flew in a Maths professor from MIT to verify some claim, so don't just accept things blindly.
Of course, none of this prevents this guy from having stolen chunks of Mozilla or something, and then bolting some bits on.
Karlin Lillington, a respected journalist for the Irish Times newspaper, maintains a weblog and has posted a more technical analysis here after talking to some people from MIT's media lab in Dublin, Ireland.
Some snippets:
"He says that what Adnan has done is re-engineer the efficiency of how a browser operates, which allows it to run up to six times faster (but usually not that much faster -- two to four times faster is more common). So it's not managing bandwidth but managing the way the browser itself handles and presents information. The researcher (whom I know and will vouch for) says that instead of simply tinkering with existing code he went down to the socket layer and reworked it at the protocol level (now, many of you guys will know the significance of this better than me, I'm just reporting the conversation). He added that it is incredibly clever work and stunning that a 16 year old has done this (I am not scrimping on the superlatives because that is what was said)."
So perhaps there is some truth in this after all.
newsQuakes