Posts by Rom Walton (BOINC)
|
61)
Message boards :
Cafe SETI :
The Ratings Game
(Message 310235)
Posted 19 May 2006 by  Rom Walton (BOINC)
Post: For those who are just as annoyed by the whole rating thing as I was, I was looking into ways to tweak it just enough that you had to have a valid RAC in order to rate. Well apparently Janus beat me to it, cuz it was already in the source tree. Later today I enabled the policy that you have to have a RAC over 10 in order to rate posts. This should keep the group of people who created a few accounts just to plus and minus people at bay. If they want to rate people from those accounts they have to pay a project tax to do it. |
|
62)
Message boards :
Cafe SETI :
Automated/Repetitive posting
(Message 310208)
Posted 19 May 2006 by Rom Walton (BOINC)
Post: Look, bottom line is I made a judgment call. Monday had no real posting history prior to the infraction and had a sizeable sum of credits which means that the account wasn’t created just to tick people off. Or if it was the person that created it was willing to throw a large amount of credit away. I disabled the account and deleted all the posts made up until that time. TBH, I really thought the account was an alternate for another user who I banned a few days before. So I didn’t even bother sending an email or anything. I made this thread sticky and made a more forceful statement about what I would do with the accounts for future reference just in case it happens again. Later that day/night I was contacted by the owner of the account apologizing for his actions and promising never to do it again. The email had valid outbound headers, so my dilemma became what was fair to the project and the user of the account? For somebody who infrequently posted and contacted me after the account was banned and wanted it back it seems to me that the right choice was to reinstate the account and see how he conducts himself. I’m still waiting for the other user who I banned a few days ago to make himself known in email or via another account. Guess we shall see what happens on that front. |
|
63)
Message boards :
Cafe SETI :
Automated/Repetitive posting
(Message 309106)
Posted 18 May 2006 by Rom Walton (BOINC)
Post: Let me state this in no uncertain terms. Automated posting will be dealt with by disabling the account and deleting all posts by said account. Period. |
|
64)
Message boards :
Cafe SETI :
Automated/Repetitive posting
(Message 305179)
Posted 14 May 2006 by Rom Walton (BOINC)
Post: Don't do it. End of story. |
|
65)
Message boards :
Number crunching :
BOINC 5.4.9
(Message 303094)
Posted 12 May 2006 by Rom Walton (BOINC)
Post: Shesh, just when you think things are going well something has to surprise ya. Okay, fixed that problem, next.... |
|
66)
Message boards :
Cafe SETI :
Star Trek News
(Message 301259)
Posted 9 May 2006 by Rom Walton (BOINC)
Post: Crystalize, your picture just cracks me up. I know this is really off-topic, but I can't help it. It reminds me so much of "Crazy Eddie" from "The Mote in God's Eye." One of the best Sci-fi books I've ever read. Without the donations S@H would close up shop. |
|
67)
Message boards :
SETI@home Science :
MAP of SETI ground covered?
(Message 274485)
Posted 2 Apr 2006 by Rom Walton (BOINC)
Post:
Well the issues above are SETI@Home specific issues, not BOINC issues. I will however point out that classic did not start out with having all though plug-ins, it was people in the community who created them. If you really need them to feel good about using the software, come back in a couple of years when the community has ported all the old classic plugins to the new BOINC based infrastructure. As for the couple of extra questions, we have something in the works for that. It'll help out with all the projects, not just SETI@Home. At the end of the day, BOINC was born to benefit all of the distributed computing projects; past, present, and future. Now there is a framework that any scientist can plug into to get some computing resources that would have normally cost a few million a year on a super computer or mainframe. All they have to do is sell the community on it. Participants no longer have to feel obligated to crunch for just one project or another, now they can crunch for as many as they deem worthy. |
|
68)
Message boards :
SETI@home Science :
MAP of SETI ground covered?
(Message 273295)
Posted 1 Apr 2006 by Rom Walton (BOINC)
Post: CRAPPY BOINC Define what you think is crappy and I'll see if I can do something about it. Unless you can define the problem space, I can't come up with a possible solution. |
|
69)
Message boards :
SETI@home Science :
MAP of SETI ground covered?
(Message 271800)
Posted 30 Mar 2006 by Rom Walton (BOINC)
Post: I’m pretty excited. ROFLMAO |
|
70)
Message boards :
SETI@home Science :
MAP of SETI ground covered?
(Message 271796)
Posted 30 Mar 2006 by Rom Walton (BOINC)
Post: The survey of the sky is probably the same, all the data only comes from one telescope. The computational engine that is currently BOINC is over twice the size of what classic was at the time it was shutdown. It is growing by a few thousand machines a day. IIRC classic was floating around 150 TeraFlops of computational capacity, but somebody who is more familiar with the classic numbers would need to speak up. As of today BOINC across all of the projects is floating around 396 TeraFlops. I looked up the stats from here. Distributed Computing needs to be more than just each project running in isolation and being setup in such a way that the community feels as though it has to choose one project over another. People have diverse interests and generally want to explore several projects. BOINC provides compromises for both the community and the project administrators so that everybody can benefit. With one client any member of the community can contribute to any project they desire. Projects get an infrastructure that can scale up to what they can sell to the community. New projects do not need to develop a complete infrastructure on there own, they get one that works and doesn’t cost them anything. Since the cost of entry for new projects is quite a bit less than before when they had to develop a complete infrastructure many more groups are willing to start up a new project. Proof can be seen in just how many new projects have started up in the last two years. Several things are in the works to address many of the usability issues the client and server software face today. I think this year we are going to see a great many changes to address the usability gap and with that we’ll see a huge increase in additional hosts. I think we are going to hit 800 TeraFlops this year, maybe even more. I’m pretty excited. |
|
71)
Message boards :
Cafe SETI :
Show off your computer desk, patch, or farm...
(Message 268512)
Posted 25 Mar 2006 by Rom Walton (BOINC)
Post: Here is my workstation. Just found this thread. :) |
|
72)
Message boards :
Number crunching :
Upgrading to Vista?
(Message 263725)
Posted 18 Mar 2006 by Rom Walton (BOINC)
Post: Actually I think only the video mini port drivers were. Disk, Network, Floppy, Keyboard, Mouse, and any special purpose device were Ring 0. Here is an article that describes the video conversion from 3.51->4.0: http://www.microsoft.com/technet/archive/ntwrkstn/evaluate/featfunc/kernelwp.mspx?mfr=true Inside Windows NT, Helen Custer, 1993, Pages 241-283. NT I/O Manager describes the nature of how NT interacts with hardware devices. Page 243 explictly describes things the keyboard, Mouse, Floppy Drive, Tape Drive, Hard Drives, Network adapters as all being kernel mode drivers interacting with the I/O manager. |
|
73)
Message boards :
Number crunching :
Upgrading to Vista?
(Message 263604)
Posted 17 Mar 2006 by Rom Walton (BOINC)
Post:
Actually I think only the video mini port drivers were. Disk, Network, Floppy, Keyboard, Mouse, and any special purpose device were Ring 0. [Edit: Actually i'll have to look that up, maybe the mini port was kernel mode, and the bulk of the driver was usermode. ] What MS found out is that even though the system stayed up if the video driver crashed, the system was unsable from the console and so most administrators still power cycled the machines. So when MS went through the process of moving the video mini port driver into the kernel they found that video performance suffered to the point of making certain graphics applications unusable, the cost of 4 user<->kernel context switches for rendering GDI was to great. So they moved a good chunk of the GDI into kernel mode. |
|
74)
Message boards :
Number crunching :
Upgrading to Vista?
(Message 263523)
Posted 17 Mar 2006 by Rom Walton (BOINC)
Post: According to the latest Singularity video the device drivers are not alloud to access memory directly, they have to go through an abstraction layer. Coupled with the fact that the device drivers have to go through an MSIL to native code conversion they can enforce at compile time that no unsafe code (pointers) are used by the driver. Like I had said, pointers are the key, if you have access to them in the kernel then you can't control access to anything. |
|
75)
Message boards :
Number crunching :
BOINC files infected with PartyPoker
(Message 263216)
Posted 17 Mar 2006 by Rom Walton (BOINC)
Post: I think CounterSpy is wrong. I think it is a false positive. |
|
76)
Message boards :
Number crunching :
BOINC files infected with PartyPoker
(Message 263203)
Posted 17 Mar 2006 by Rom Walton (BOINC)
Post: Actually, that file is used as a file compression library. It has nothing to do with party poker. I can assure you of that. I built it from the zlib sources myself. |
|
77)
Message boards :
Number crunching :
Upgrading to Vista?
(Message 263172)
Posted 17 Mar 2006 by Rom Walton (BOINC)
Post:
It is being done in Vista. Instead of the driver being in ring 0 though it was moved into Ring 3. It is for only a certain classification of drivers though.
Okay, I think I know where our disconnect is here. When the call from user mode is made to switch to kernel mode happens, the executing security context is also changed from the context of the user executing the application to SYSTEM. Everything that runs in the kernel runs as SYSTEM. Now the kernel mode code can lookup the security context of the process it is supposed to do something with, but it doesn't have too. It is already running as the highest authority on the computer. If the kernel were to lock itself out of a certain block of memory, it wouldn't be able to get access to it again until the machine is rebooted. So applying ACLs to kernel mode objects for kernel mode code is kind of pointless since you can go to any point of memory you want and call any function you want, even ones with no names or no published function prototype. In the end, it is all about the pointers and being able to walk through memory one byte at a time that keeps the whole acl for kernel mode objects for kernel mode code at Ring 0 problematic. Anyways, certain drivers are being moved to Ring 3. So they'll no longer be a problem. Back to work. |
|
78)
Message boards :
Number crunching :
Upgrading to Vista?
(Message 263104)
Posted 17 Mar 2006 by Rom Walton (BOINC)
Post:
If I remember correctly all IIS does for CGI applications is add environment variables for each piece of data defined by the CGI specification. So your CGI application could be a batch file, perl script, php script, or any other language you want to use. ASP.NET is actually an ISAPI application. Look for: C:\\WINDOWS\\Microsoft.NET\\Framework\\v2.0.50727\\aspnet_isapi.dll The ISAPI filter looks at the web.config file in the root of the project folder to determine which DLL's to load up and which pages need to be just-in-time compiled. The newly compiled pages and binaries are loaded up into an w3wp.exe (win2k3) process or aspnet_wp.exe (WinXP). ISAPI's were first created to elminate the overhead of process creation. Process creation is expensive. It gets even work when you start trying to create 500-1000 processes a second. ISAPI's biggest draw back is that they were loaded up as part of the IIS process space. If the ISAPI had a bug or leaked memory eventually IIS would crash.
Mach-O had the same conceptual model with the exception being that it wasn't managed code. Managed code is nice, but you pay a 10-15% performance cost for it. |
|
79)
Message boards :
Cafe SETI :
How geek'ish are you?
(Message 261652)
Posted 14 Mar 2006 by Rom Walton (BOINC)
Post: I was rated at 42.99803% - Major Geek. Woo Hoo. |
|
80)
Message boards :
Number crunching :
Upgrading to Vista?
(Message 260860)
Posted 12 Mar 2006 by Rom Walton (BOINC)
Post:
What process? A process is a virtual address space, in the kernel you see the values and locations of the machines physical memory.
Whether you publish a function or not, it can still be called via a kernel mode driver. All you need is the function pointer, and all you need to get that is to hook up a kernel debugger. Another tid bit. Public, private, and protected class modifiers are only used during compile time. At runtime you can still modify any value in a class at anytime, all you need is a pointer to the class and the offset of the variable you want to modify. As far as this conversation, it needs to be stopped or taken offline. |
Previous 20 · Next 20
©2020 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.