I dont believe in public networks of distributed computations.
First, because desire of participants to make the way on top of a rating or, especially, to receive his copeck is mach above their desires to execute the purposes of the project, it is known and obvious fact.
And secondly, because "noncommercial" structures on maintenance of the infrastructure of the distributed calculations, such as dnet, also are has little interest in performance of the project as at present they operate as dotcoms in their time - "we shall stake out a place in the market, and then we shall think about protection and everything else".
But I am deeply confident, as after the period of initial formation of the market, they will not engage it, for the simple reason: application of more or less adequate measures of protection reduces efficiency of the calculator, that in conditions of a competition reduces appeal in opinion of the customer. And it is known, that for the company it is usually easier to sweep dust under a carpet, than to call a command of ashmans, pay them a part of the profit, and in every possible way to strain. In  dnet have shown, that they in general know about means of protection in distributed computations, however for some reasons do not apply them. The witty reader, probably, will comment, that they may apply, only I do not know about it. On that I at once want to ask: why, then their system the second month accepts my dummies, displays me in ratings, and in general reports about a gain of activity in their project? In the following section receptions which I used are described. For ignorant I shall note, they can be applied by anyone who has read one clause about "How to become a hacker" :), the big knowledge and experience in reverse engineering are not required - it takes 20 minutes in the evening after work. Is there somebody to convince me, that I was the first who have guessed?
Thats the why I find it dangerous to trust open systems of the distributed computations projects of any social importance - we cannot trust their results, we cannot trust them at all. However, and that is more dangerous, it will be hard to deny such a results just because nobody can verify them.
Certainly, it would be silly to reject the idea of distributed computations. Fruitful work on their development and use is conducted both in Russia  and in other countries, and these projects bring real advantage. However, these are the systems of controllable computing resources, where participants are more interested in useful result than in high position in the generated html-table. But for systems such as dnet, on their today's level of development, I do not find a place.
I am not a member of any team participated in dnet projects. I investigated dnetc only for educational purposes.
The main idea of the tweak is to skip computational block of the algorithm, The client is still honestly search keys, but does not calculate them by RSA. Thats the source not that large increase in speed - only in 66 times, but it is makes output blocks a little more credible.
As example, I alter the "SS 2-pipe" core. Thus you can see a difference only if you switch to this core. The tweak can be applied to other cores as well. There is little technical information to talk about. It is just a vulgar jump-fix - the lowest of the crackers activities, the one that everyone can do.
For wise ones, who doesnt download and execute any and every executable they find in the Internet (if you not of such, you just a stupid) I tell why it is not dangerous in this case.
The official dnet client version 2.9005.483 for Windows 32bit is compressed with UPX (the Ultimate Packer for eXecutables). If you decompress it "upx -d dnetc.exe" you will get file of the same size as tweaked one from this page. If you then compare them, you will see 5-byte difference - unconditional jump. In another words, it is neither a virus, nor a trojan horse, nor a block redirector to my address. But I still have no responsibility for any damage this tweak can or cannot produce.
This tweak is present only for show how it is easy to cheat in dnet.
You can freely download this tweak. But ONLY for testing and educational purposes. It is neither legal, nor right to use such things for any other purposes. Be warned also, that using this tweak for "computing" real blocks and send them to the root server, you will expose not only yourself, but also your team.
core=5line to "[rc5-72]" section in [dnetc.ini] (if there is no such section - create it).
Q. Do you think this page is written in English?
A. No, I dont think so :).
Q. Is it still working?
A. 24-Sept-2003. This issue is still valid.
Q. Why do you do all this?
A. It is interesting.
Q. Do you want to compromise all the work that thousands of the people all around the world done so far?
Q. Do you want to halt the dnet's RC5-72 project?
Q. Are you connected to the "Cult of Dead Cow" :).
Q. What is "Tlo"?
A. The set of letters. There are Russian words "zlo" (evil) and "tlen" (decay), and they have nothing common with tlo :). Because Tlo's alignment is "chaotic-good".
. Operational Code Authentication. (c) Jeff Lawson 2000. [http://www.distributed.net/source/specs/opcodeauth.html]
. Суперкомпьютер на выходные. (c) Открытые системы. 2003. [http://www.osp.ru/os/2003/05/043.htm]
. Uncheatable Distributed Computations. Philippe Golle and Ilya Mironov. [http://crypto.stanford.edu/~pgolle/papers/distr.html].
Contact me via "tlo AT inbox PERIOD ru".