Except for the detail that the current version of mldonkey leaks memory like a sieve -- it will eat hundreds of megs of RAM after running for a while. And of course, ocaml has jack shit by way of memory profiling tools, so despite the fact that it's a garbage collected language and *theoretically* avoids memory problems there's masses of crap accumulating somewhere. Probably a hash table somewhere. In a C program, a memory profiler would have turned up the problems almost immediately. In an ocaml program, people just bite their lip and suffer and restart the program periodically.
They will if someone complains about a specific instance.
Should your ISP have blocked your post because you talked about underage porn?
No. My ISP is not acting as a content provider, nor am I providing underage porn, just discussing it.
On the other hand, go to a forum that allows the posting of images (say, the fark forums), and post an IMG reference to underage porn and you'll find that image going away damned quickly.
And it's halfway there, but by not using GPG, it means that it can't piggyback off of all the other work that I do to maintain a trust database (I'm not going to maintain a different list of identities of people that I know of for every single program I use -- that's just unreasonable.) Also, it isn't a standard -- nothing interoperates with it. Sure, licq has a method of encrypting messages (actually, might just be SSL instead of full end-to-end encryption -- I'd have to look), gaim has two, jabber clients have at least one). None of them interoperate, so nobody uses them.
If a client did something as simple as taking a random number at the start of each session and sending it to the remote client, and then every message from the remote client had, as a header, a nonce consisting of tuple of a local random number (an increasing sequence number) and the sent-at-session-initiation random number, and then each message was GPG encrypted and signed, you have a standard mechanism that can be used by any client just by feeding the data into GPG -- a *standard* mechanism that every client can support. (The random numbers are necessary to avoid replay attacks -- else I could log someone saying "Sure, I give you authorization to do that" and then use that statement in another conversation. That's not a big deal from an encryption standpoint, but from a signing standpoint, especially as IM is being used in business now, it's serious.)
So, consulting firms like Gartner say "Windows is insecure". Big deal. Garter is for hire for PR fodder. You know who to ask if you want the real dirt on what has problems? IT professionals, the sort of people who frequent Slashdot. Garter is trying to approximate what an IT professional would say.
Do I think Windows has security problems? Sure, both in Microsoft applications and in API at a design level. There is also some missing security functionality, like a sandboxing mechanism. However, I think more of the problem comes from a long tradition of single-user systems and application developers not writing security-conscious code. Who calls out Adobe for, say, opening a local system vulnerability with Photoshop? Nobody. On the other hand, if OSS/Linux or Oracle opens a hole on a *IX box, then people make noise.
My issue is not that Microsoft is accused of having security problems when they don't have any (though, to be fair, Linux isn't perfect either). No, my problem is that *Gartner* saying that Microsoft is insecure should mean nothing to a typical Slashdotter. A typical Slashdotter should be relying on their own experience, not on Gartner. Gartner is for large company CIOs, suits that don't understand technology and want their business decisions fed to them ground up into a nice paste.
If it was like anything around here, you get fingerprinted, and they hand the fingerprint card to your parents in case you're kidnapped. The police don't want a bunch of toddler prints lying around.
Why not? Your prints don't change shape. You just slap them in a police database. And while I was a kid, I certainly wasn't a toddler.
I'd prefer that it be a CGI argument (or at least the option of a CGI argument) to disable news.
I'm very uncomfortable with the idea of Google building a profile on me, and the idea of "preferences", which entails the use of cookies. If it's a CGI argument, I can just save a bookmark with the preferences I want. If it's a cookie, I have to set my preferences each time, and there is a strong disincentive to maintain privacy.
Given China's behaviour with respect to Tibet and Taiwan, I would say that any company that specifically re-enforces the policy of the government through censorship has no more right to claim to not be evil than Fox News has to claim to be fair and balanced.
So consider the case of underage pornography (something that the US government does censor). Should Google not censor it?
All governments that I know of do *some* censorship -- the question is just to what degree.
I mean, I think that the people running China are a bunch of shortsighted assholes, but they aren't qualitatively different from other governments -- just, perhaps, quantitatively. Given that we listen to US media, we hear a lot about how awful China is doing.
On the other hand, the US Iraq occupational authority did not allow freedom of press, and in fact shut down a number of media sources for criticizing them (newspapers and the only Arab-language news network). Naturally *that* didn't get much air time -- but godless communist oppressors censoring critical media is acceptable and *required* content for us to hear about.
West Virginia does not require you to thumbprint when you get your driver's license, but they do not inform you of the fact at the time.
I *was* fully fingerprinted (i.e. not just thumb) by the local police department when I was a child as part of some elementary school-related function. I remember objecting, even at that age, and being told that it was "in case I was ever kidnapped." I was kind of dubious, and tried smudging them.
Because once you have lost data privacy, you're never, ever going to get it back.
One more database falls to the federal government.
I can't wait until the first person prosecuted or watch-listed because of something he said over an instant-messaging program ("God, Bush is an idiot -- I wish someone would shoot him".) Still no GPG encryption on IM clients (well, other than gabber).
Used to be that you could have an anonymous website, but that's about to go away.
You can't drive without a license (where you get thumbprinted).
You can't fly without all sorts of data about you being logged.
The US government is pushing hard for biometrics in all areas. Biometrics are *terrible* as a traditional authentication system mechanism, since once someone's stolen the secret data (say, hacked one iris reader), you can never invalidate it. However, they're wonderful for monitoring purposes, since people have their "papers" with them wherever they go. They can also be used to tie together databases nicely.
Authoritarianism allowed by the application of computers will be one of the greatest new world problems that we'll have to face. Never before have societies had the ability to crack down, monitor, and ensure precisely compliant behavior on such a large chunk of their population. Can humans function well in such an environment? Is such an environment a good idea?
Well, up to five years, not a guaranteed five years. 3.a.2.B lists the "illicit authentication feature" that I'm referring to, and 3.a.3 lists the five year prison penalty.
3.6.A has a definition of what an illicit authentication feature is.
I output a frame, start decoding a new one to fill my buffer again... and after 1/25 seconds I do what again if ffmpeg isnt finished yet? I wait that is what I do if Im single threaded.
You save state, slap the frame up, and resume decoding.
With a multithreaded implementation the OS would schedule an output thread, put up one of the frames from the buffer and retain smooth playback.
With a multithreaded implementation, the OS does some of the work. The OS is a general-purpose OS and until recently had latencies causes serious issues with real-time things like sound work. It has no idea what kind of latency you require, what work should be done first, or anything of that nature.
You can stay in sync just easily with a single threaded implementation, but you cannot maintain the same fluidity.
You certainly can. You just have a do_pending_decoding_work() function that returns after a maximum amount of time instead of a decode_one_frame() function.
We have two bills -- one of the two linked-to ones makes writing software like Daemon Tools, no-CD patches, and bnetd (a libre open source battle.net implementation) a federal crime.
The second bill makes not providing identifying information to let the feds track you down if they know what your website is illegal. (What if I want to speak freely, without fears of being harassed? I can post papers anonymously, but not be anonymous on the Web?) Add in the next obvious thing, a requirement for webmasters to log and be able to provide information for who posted something, and federal law enforcement can track anyone down.
Combine this with the fact that Cat Stevens just fell under the eye of the Homeland Security Watch List, had his plane diverted to Maine and was kicked out of the United States. As far as I can tell, his main crime was criticizing US involvement in Iraq.
And Bush's polls are looking better than ever.
It's an authoritarian next few years for all of us...
The Federal Anticounterfeiting Act of 2004, the bill that was *actually* linked to, is some scary stuff.
Writing a program like Daemon Tools (no, not the *IX suite of software, the CD image software) or bnetd (a FOSS Battle.net implementation) would become illegal, with a potential five year federal prison penalty.
I'm not real sure that I want some MCSE patching the system that tracks the plane that I'm in. Sure, you don't *think* that there will be any side effects...
Slashdot Mirror
If mldonkey's memory leaks ever get removed, it might be usable.
As one of the most widely used OSS showcase apps for ocaml, mldonkey is a bit of an embarassment.
Except for the detail that the current version of mldonkey leaks memory like a sieve -- it will eat hundreds of megs of RAM after running for a while. And of course, ocaml has jack shit by way of memory profiling tools, so despite the fact that it's a garbage collected language and *theoretically* avoids memory problems there's masses of crap accumulating somewhere. Probably a hash table somewhere. In a C program, a memory profiler would have turned up the problems almost immediately. In an ocaml program, people just bite their lip and suffer and restart the program periodically.
Of course google shouldn't censor underage porn.
They will if someone complains about a specific instance.
Should your ISP have blocked your post because you talked about underage porn?
No. My ISP is not acting as a content provider, nor am I providing underage porn, just discussing it.
On the other hand, go to a forum that allows the posting of images (say, the fark forums), and post an IMG reference to underage porn and you'll find that image going away damned quickly.
And it's halfway there, but by not using GPG, it means that it can't piggyback off of all the other work that I do to maintain a trust database (I'm not going to maintain a different list of identities of people that I know of for every single program I use -- that's just unreasonable.) Also, it isn't a standard -- nothing interoperates with it. Sure, licq has a method of encrypting messages (actually, might just be SSL instead of full end-to-end encryption -- I'd have to look), gaim has two, jabber clients have at least one). None of them interoperate, so nobody uses them.
If a client did something as simple as taking a random number at the start of each session and sending it to the remote client, and then every message from the remote client had, as a header, a nonce consisting of tuple of a local random number (an increasing sequence number) and the sent-at-session-initiation random number, and then each message was GPG encrypted and signed, you have a standard mechanism that can be used by any client just by feeding the data into GPG -- a *standard* mechanism that every client can support. (The random numbers are necessary to avoid replay attacks -- else I could log someone saying "Sure, I give you authorization to do that" and then use that statement in another conversation. That's not a big deal from an encryption standpoint, but from a signing standpoint, especially as IM is being used in business now, it's serious.)
So, consulting firms like Gartner say "Windows is insecure". Big deal. Garter is for hire for PR fodder. You know who to ask if you want the real dirt on what has problems? IT professionals, the sort of people who frequent Slashdot. Garter is trying to approximate what an IT professional would say.
Do I think Windows has security problems? Sure, both in Microsoft applications and in API at a design level. There is also some missing security functionality, like a sandboxing mechanism. However, I think more of the problem comes from a long tradition of single-user systems and application developers not writing security-conscious code. Who calls out Adobe for, say, opening a local system vulnerability with Photoshop? Nobody. On the other hand, if OSS/Linux or Oracle opens a hole on a *IX box, then people make noise.
My issue is not that Microsoft is accused of having security problems when they don't have any (though, to be fair, Linux isn't perfect either). No, my problem is that *Gartner* saying that Microsoft is insecure should mean nothing to a typical Slashdotter. A typical Slashdotter should be relying on their own experience, not on Gartner. Gartner is for large company CIOs, suits that don't understand technology and want their business decisions fed to them ground up into a nice paste.
If it was like anything around here, you get fingerprinted, and they hand the fingerprint card to your parents in case you're kidnapped. The police don't want a bunch of toddler prints lying around.
Why not? Your prints don't change shape. You just slap them in a police database. And while I was a kid, I certainly wasn't a toddler.
Sorry. That should be "disable subscription news", not "disable news".
I'd prefer that it be a CGI argument (or at least the option of a CGI argument) to disable news.
I'm very uncomfortable with the idea of Google building a profile on me, and the idea of "preferences", which entails the use of cookies. If it's a CGI argument, I can just save a bookmark with the preferences I want. If it's a cookie, I have to set my preferences each time, and there is a strong disincentive to maintain privacy.
Given China's behaviour with respect to Tibet and Taiwan, I would say that any company that specifically re-enforces the policy of the government through censorship has no more right to claim to not be evil than Fox News has to claim to be fair and balanced.
So consider the case of underage pornography (something that the US government does censor). Should Google not censor it?
All governments that I know of do *some* censorship -- the question is just to what degree.
I mean, I think that the people running China are a bunch of shortsighted assholes, but they aren't qualitatively different from other governments -- just, perhaps, quantitatively. Given that we listen to US media, we hear a lot about how awful China is doing.
On the other hand, the US Iraq occupational authority did not allow freedom of press, and in fact shut down a number of media sources for criticizing them (newspapers and the only Arab-language news network). Naturally *that* didn't get much air time -- but godless communist oppressors censoring critical media is acceptable and *required* content for us to hear about.
West Virginia does not require you to thumbprint when you get your driver's license, but they do not inform you of the fact at the time.
I *was* fully fingerprinted (i.e. not just thumb) by the local police department when I was a child as part of some elementary school-related function. I remember objecting, even at that age, and being told that it was "in case I was ever kidnapped." I was kind of dubious, and tried smudging them.
Because once you have lost data privacy, you're never, ever going to get it back.
One more database falls to the federal government.
I can't wait until the first person prosecuted or watch-listed because of something he said over an instant-messaging program ("God, Bush is an idiot -- I wish someone would shoot him".) Still no GPG encryption on IM clients (well, other than gabber).
Used to be that you could have an anonymous website, but that's about to go away.
You can't drive without a license (where you get thumbprinted).
You can't fly without all sorts of data about you being logged.
The US government is pushing hard for biometrics in all areas. Biometrics are *terrible* as a traditional authentication system mechanism, since once someone's stolen the secret data (say, hacked one iris reader), you can never invalidate it. However, they're wonderful for monitoring purposes, since people have their "papers" with them wherever they go. They can also be used to tie together databases nicely.
Authoritarianism allowed by the application of computers will be one of the greatest new world problems that we'll have to face. Never before have societies had the ability to crack down, monitor, and ensure precisely compliant behavior on such a large chunk of their population. Can humans function well in such an environment? Is such an environment a good idea?
Well, up to five years, not a guaranteed five years. 3.a.2.B lists the "illicit authentication feature" that I'm referring to, and 3.a.3 lists the five year prison penalty.
3.6.A has a definition of what an illicit authentication feature is.
I output a frame, start decoding a new one to fill my buffer again ... and after 1/25 seconds I do what again if ffmpeg isnt finished yet? I wait that is what I do if Im single threaded.
You save state, slap the frame up, and resume decoding.
With a multithreaded implementation the OS would schedule an output thread, put up one of the frames from the buffer and retain smooth playback.
With a multithreaded implementation, the OS does some of the work. The OS is a general-purpose OS and until recently had latencies causes serious issues with real-time things like sound work. It has no idea what kind of latency you require, what work should be done first, or anything of that nature.
You can stay in sync just easily with a single threaded implementation, but you cannot maintain the same fluidity.
You certainly can. You just have a do_pending_decoding_work() function that returns after a maximum amount of time instead of a decode_one_frame() function.
IPv6 already specifies a range of addresses that map to IPv4 addresses.
Has anyone ever noticed that what everyone *really* wants is a sex chatbot?
We have two bills -- one of the two linked-to ones makes writing software like Daemon Tools, no-CD patches, and bnetd (a libre open source battle.net implementation) a federal crime.
The second bill makes not providing identifying information to let the feds track you down if they know what your website is illegal. (What if I want to speak freely, without fears of being harassed? I can post papers anonymously, but not be anonymous on the Web?) Add in the next obvious thing, a requirement for webmasters to log and be able to provide information for who posted something, and federal law enforcement can track anyone down.
Combine this with the fact that Cat Stevens just fell under the eye of the Homeland Security Watch List, had his plane diverted to Maine and was kicked out of the United
States. As far as I can tell, his main crime was criticizing US involvement in Iraq.
And Bush's polls are looking better than ever.
It's an authoritarian next few years for all of us...
The Federal Anticounterfeiting Act of 2004, the bill that was *actually* linked to, is some scary stuff.
Writing a program like Daemon Tools (no, not the *IX suite of software, the CD image software) or bnetd (a FOSS Battle.net implementation) would become illegal, with a potential five year federal prison penalty.
Why have I not heard about this before?
a) That's nonportable. Use telinit 6.
b) What, you don't like "reboot"?
Aviation expects 99.9999% uptime with absolutely no message loss, and we would achieve that with hot-standbys and MySQL mirroring.
;-)
Yes, that was a jab at you, Postgres fans.
I'm not real sure that I want some MCSE patching the system that tracks the plane that I'm in. Sure, you don't *think* that there will be any side effects...
True, but the also-POSIX-compliant time() is going to cause plent of interesting problems in Y2038.
[shrug] I use two scripts:
/, "\\ ");
#!/bin/gawk -f
# bashescape.awk
{
gsub(/\\/, "\\\\");
gsub(/
gsub(/!/, "\\!");
gsub(/"/, "\\\"");
gsub(/'/, "\\'");
gsub(/:/, "\\:");
gsub(/;/, "\\;");
gsub(/=/, "\\=");
gsub(/?/, "\\?");
gsub(/@/, "\\@");
gsub(/\^/, "\\^");
gsub(/{/, "\\{");
gsub(/}/, "\\}");
print;
}
#!/bin/bash
# myxargs
bashescape.awk|xargs "$@"
The problem is that xargs isn't a shell builtin, so it doesn't really know what the characters you need to escape for your shell is.
The latter script, the xargs replacement, is tremendously useful.
-vo xv would normally go in your config file.
And I lied -- it should be -fs, not -fs=yes.
That's just the same as hitting "f" to fullscreen the movie.
So I have vo=xv and fs=yes in my config file, and don't have to worry about it. It's covered in the docs.
A case of bad in theory good in practice.
Broken theory.
(though it does take a while to resync)
See the mplayer man page:
-mc
Maximum A-V sync correction per frame (in seconds).