Slashdot Mirror
Bram Cohen's Response to Microsoft's Avalanche
An anonymous reader writes "Bram Cohen has reduced Microsoft's proposed file-sharing application--codenamed Avalanche--to vaporware, dubbing its paper on the subject as "complete garbage". "I'd like to clarify that Avalanche is vapourware," Cohen said. "It isn't a product which you can use or test with, it's a bunch of proposed algorithms. There isn't even a fleshed-out network protocol. The 'experiments' they've done are simulations.""
The question is: is it GOOD vaporware? Can the proposed algorithms deliver the results we want? That Microsoft has the manpower to turn it into real software is a given.
The illegal we do immediately. The unconstitutional takes a little longer.
--Henry Kissinger
RTFA! It's still there people! But if you haven't RTFA and want to spout off something please keep in mind the following: He invented Bittorrent, and his biggest problem is this "research" is based on massive misunderstandings of how Bittorrent actually works.
Actually, even as a research paper, the paper falls rather short. There has been prior work on P2P using forward error correction (FEC). FEC also means that you don't have to get every single chunk, but that you can reconstruct missing chunks from data you already have.
The authors should have demonstrated that their approach is better than FEC-based P2P protocols, but instead, they only compared it to simple P2P protocols. So, their protocol may actually not be better than the state of the art at all, and may actually be harder to implement in practice.
Yup. Bram's blog makes it clear that MS's approach to P2P is to reverse engineer an outdated version of his protocol. The result will probably be similar to MS's effort to reverse engineer the Macintosh: Windows 3.1 -- meaning that it will be buggy, slower and vastly more popular.
Keep reading Bram's blog. How far do you get before it starts going over your head? The dude has skillz that dust 98% of the wannabes here on Slashdot.
And as for motives, in my experience with autistics, it's common for those with Asperger's Syndrome to be quite guileless. They speak and act without consideration for other's "feelings". As a result they are more frank and honest than most people are comfortable with. Sorta like if Mr. Spock insults your work. He's not doing it to hurt you, or out of jealousy, he's saying it because it is the most logical observation.
If the originating server has to calculate the hashes, then it would have had to calculate the FEC blocks that the hashes are calculated for as well.
Ergo, the network coding advantage is lost.
Unless there is some way to compute hashes of the FEC blocks without actually haveing the blocks themselves, there would be no advantage of Avalanche over a server-implemented Tornado code..
I'm starting to see Avalanche as only becoming used as being used within a `locked' setting. Probably something that'll come shortly after Longhorn and Trusted *** start infiltrating the MS desktop.
argumentum ad fallacium: Fallacy of defining a fallacy which allows one to dismiss the argument in question.
i have my doubts about some (but not all!) of the research that people undertake in the academic world of computer engineering/systems. from what i have seen, some academics will rush to publish for the sake of having a paper, even if it means cutting corners on an otherwise interesting idea or application. while i won't say that this is the norm at the very best institutions in the world, things on the whole can seem less than perfect.
h tm
i imply nothing about the individuals in the paper that Bram attacks, since i haven't interacted with them firsthand. however, it might be interesting to note that the primary author is a grad student at Georgia Tech. according to his web page, his stint at MS research was just a ~6 month period, 2/04-6/04 & 7/04-8/04:
http://www.cc.gatech.edu/~gantsich/biography.htm
the call for papers for this 2005 conference set a deadline of 7/7/2004:
http://www.ieee-infocom.org/2005/call_for_papers.
this does not leave a huge block of time for one student to brush up on the research background, flesh up the practical aspects of the idea, implement (and validate?) a simulator, complete a preliminary set of data runs, and write a paper draft worthy of acceptance. let's not forget any downtime that might arise at the start of an internship (moving over the pond, getting acclimated, etc.).
here, i assume the not unrealistic situation where the official research scientist principally serves as a primary investigator. he brews the idea, perhaps working out some more theoretical aspects of the problem, and handles all the headaches related to funding/approval/propaganda. this entrusts a good deal of the grunt work to the student. i tend to see this sort of behavior in the ivory tower, but it is entirely likely that research in industry is much more balanced!
time should not be an excuse in any case, but it does raise an eyebrow toward the paper-happy nature of some research these days. you make the call on what you believe is reasonable concerning those flaws in methodology that Bram has so derided in his blog.
does anyone have a clue about the timing of the media's spin on things? The Register's article from the first slashdot posting is one of the first according to Google News...