Well, the problem, to put it simply, is lack of control. Govt. is (or at least should be) bound by restrictions, i.e. police needs a warrant to monitor your conversations, and we (the public) could observe them and check that they do it for public good.
There's nothing, on the other hand, that stops corporations s.a. Verizon, from monitoring your calls for profit only, uncontrollably.
Secodndly, the government is completely different from this. It exists to advance a political agenda and control every detail of our lives. It has a moral outloook, and if your morals are different you are screwed.
You elect the government. Dont like police beating - vote Harry Browne. Want clean air - vote Ralph Nader. Want to see how dumb a president could be - vote... oh, never mind that one...
On the contrary, you do not elect corporations. You are forced to obey whatever they tell you to do pretty much the same way you're forced to obey laws - i.e. techinically you could ignore laws or corporate agreements, but you'll face the consequences
Corporations have a record of non-abuse, and are owned by the people.
Not exactly, but I like your sence of humor. Corporations care about people, help people and protect people. For a list of people browse www.forbes.com.
Opinions are mine only and could change without notice.
You can't stop M$ from collecting your data on their servers - if it against trhe law they'll break the law, simple as that. This is not the first time M$ collects this kind of data - remember M$Office GUIDs, regwiz tricks, etc. Whatever is the reason they need this data is a mystery, but it seems like they are trying to do this repeatedly.
However, you could poison the data, i.e. make a lot of invalid entries - like creating new account on Hotmail each time you need to lookup something on MSDN, etc. As signal-to-noise ratio goes down M$ would end up with huge database, most of data in which is outdated and therefore useles for most Evil Purposes(TM).
As for secure trusted email the best I've seen is Lokmail - they support standart PGP.
Opinions are mine only and could change without notice.
I do agree that govt control is supposed to be better than corporate control (worst case scenario: all politicians are evil, govt cares about themselves only, but has to please voters to get re-elected - i.e. in worst case govt is the same as corporations).
But are there only 2 choices ? What if instead of some agency you would have several volunteer-driven groups (sites) targeted at specific audience which would link to information and provide commentary ? And where readers could then discuss it further ? And then you decide which groups you trust to classify information for you ? You pick several sources to filter "censor" information, and submit information you consider valuable.
I believe I heard about site called... umm Slashdot was it ? Or kuro5hin ?
There you could get "pre-filtered" information that could be then further corrected/discussed in comments. Anyone else around here heard of Slashdot , no ?
Opinions are mine only and could change without notice.
Re:Inefficienct but useable and can do more
on
Perl and .NET
·
· Score: 2
I agree with practically everything you say, XML is great for exchanging complex data structures with added benefit of error checking, etc. That's what XML was designed to do.
However, if I understand correctly - I can't find any exact description of what.NET internal structure looks like - M$ is using it for procedure calls. XML is a right tool for data exchange, CORBA is right for procedure calls, and I thought we're supposed to use right tool for right job, or even beter, have a choice what to use - Java doesn't force you to use CORBA, you could use XML RPC if you want - heck, it you really need you could send Java class as data and let it run on remote machine, with it's own security manager, if you want - sometimes this is the best solution for distributed systems. But Java gives you a choice. I can't find any evidence that.NET does, hopefully I'm wrong.
Opinions are mine only and could change without notice.
Well, CORBA is not "impossible for developers to use". See here, you create object interface definition using IDL - which isn't any harder than creating XML DTD - and then it is compiled by IDL com piler in binary form. Easy and efficient. Now SOAP seems to - correct me if I'm wrong - at least involve XML parsing, sending data in uncompressed text format, extracting SOAP protocol data from parsed XML query - and all this is encapsulated in HTTP query. Lot's of processing, for, take a note - each SOAP object call. It could be optimized, I suppose - persistent HTTP connection, maybe compressed data for bandwith - but then it's uncompressing overheat, etc - but CORBA does this already.
My guess is sooner or later we'll get binary XML encoding, then some "preparsed SOAP", and then finally something similar to CORBA - or M$ would switch to CORBA, which is unlikely because it means, gods forbid, Java compatibility
Opinions are mine only and could change without notice.
Could I legally get the description of source code from, say, some anonymous person in Russia, and then make clean-room implementation ? i.e.
h@x0r:then it opens "win.ini" for write... antv:C-x win.c if((fi=fopen("win.ini","a"))!=NULL) C-x *zenirc* Checkski h@x0r:.. and writes "shell=command.com" in there antv:C-x win.c fprintf(fi,"shell=command.com\n"); C-x *zenirc* Checkski
Legally speaking I never seen the code, so is this legit ?
Umm, any reason why we need such military power in a first place ?
Soviet Union is gone, US is the biggest nuclear power on this planet, no enemies to defend from,
at least with Army - to defend from all those
terrorists that exist according to current Govt., you would need law enforcement, not Army.
So is there any reason for strong military other than justifying budget spending ?
I guess it's mostly support issue - they don't know how or don't want to fix Linux problems.
At my current work they gave me machine with NT 4 taking 4 gigs and 9 gigs of empty space. They said I could install anything I want, but they would only support NT and, surprisingly, FreeBSD up to some extent. I decided to go for BSD since I wanted to learn it. So I fixed NT in a usual fassion (mkfs) and had no problems with IT anymore. I also have Linux on laptop, use it in company. I know that almost every tech in my company uses either Linux or BSD, or Solaris for those with Sun boxes, graphic designers use Mac. The only thing they require is for you to be compatible (SAMBA to access printers, have to be able to read.DOC, etc). And officially IT is not supposed to help you if your Linux box is screwed - I know my bos got about 5 different Linux distros from IT when he asked for it - but officially they aren't responsible. Beyong that - who cares what you use as long as you do your work ?
On my previous work, when I was admin myself, I used Linux as universal troubleshooting tool - had it work as print server for Macs and reroute all print jobs to Netware server, etc. Had to fix couple of linux boxes for professors - it was a college work - I know that most other admins never bothered to fix Linux, or for that matter anything that they had any excuse not to fix. That's probably why some IT managers dislike Linux or anything else that is new - they don't know what security measures are needed for system, how to support it, would it interfere with anything (one of my friends used Linux with Netware, named machine/share or smth like that using Russian characters, half of their windows boxes started crashing when trying to browse network).
StegFS allows you to have mandatory number of data layers. w/o knowing password there is no way to determine how many layers are there. They ask for password, you give them one. Or two. Or five. There is no way to determine if you had given them passwords to all layers of data.
M$ is mixing together several unrelated things, as usual.
Eye tracking technology While computer noticing user attention is not such a bad idea, does it needs to be bundled with UI ? It might be useful in some programs, not so useful in other. Having browser to automatically scroll down page as I read it is cool. Having word processor reposition my cursor every time I look at previous sentence, pop up dialog box while I'm using photo editing or CAD program, launch calendar each time I look at clock icon - all this is just about as useful as having paperclip pop up saying "You appear to be writing a letter"/* I realise M$ software is tested on M$ employes, but I'm not that stupid (so I use Linux at home, FreeBSD at work) */. Jokes aside, it would be good to have something like an optional library for "vision input" that software could use. It would be disaster if this thing would be mandatory bundled with UI.
UI stealing focus. M$ UI is very nasty to say the least. The most annoying part is that they don't have virtual desktops as default (there are 3rd party hacks tho). They also don't have any way to customise how and when your windows are raised, etc. I use Sawfish (aka Sawmill) as my WM, it has features like shade-hover, i.e. window is shaded but becomes unshaded after I point mouse on it, auto-raise after some time windows are raised up, window grouping i.e. EveryBuddy windows (AIM/ICQ/etc) are raised on top, but are auto-shaded and don't get focus by default, so if someone IMs me I notice it but don't get distracted. All those features help you to reduce amount of unwanted info. Also, Gnome have incredibly useful feature - tear-off menus. That way you could keep all things you need around. That's the kind of UI that makes it easier to focus on work (or Slashdot) and don't be distracted. And, the most important part, you customize it for yourself, so it doesn't give you any trouble (in Sawmill you could even write macros for WM (like animated desktop sliding (or any other kind of actions/customizations (Sawmill have built-in lisp interpreter (and since it's LISP you could very easily write AI programs;-) (as well as anything else - LISP is very powerful (some people even say it changes the way you think (but that's not true, of course)))))))). M$ doesn't have that kind of UI. That's one thing they definitely should implement - user-customizable UI - before adding any features - be it voice-recognition, eye tracking, or whatever
So, again, this technology is kind of cool, but it's nothing more (or less) than new input method, and it should be up to software developer to decide wether his/her particular application needs it.
P.S. Almost forgot - M$ sucks no matter what !!!!!
Well, this seems to be a problem with any anonymous multiuser information system - attacker could try to disrupt service by posting "noise" data. Since there is no algorithm which could validate wether the file is noise there should be some kind of moderation system. For example Napster/Gnutella/whatever client generating CRC check of file and chatrooms posting correct CRC. Napster should probably implement "preview seek" feature to listen to random part of song - since people usually remember songs by words in a middle, not by intro, anyway. Another feature I really miss in Napster is download resume - because of that there are so many truncated songs - sort of noise, too. Gnutella has this feature. As I said, CRC sums of songs and something like chatrooms with "moderators" providing correct CRC, and people voting would probably cut down noise level. That's similar to what Slashdot does - coments arent's thrown out, but some get higher rating.
Fundamental problem with any digital media protection system is that it must produce decrypted output for the end user. This means even if Big Brother Records Inc. would require me to use special device attached to computer, that decodes music according to my fingerprints, sound output would still go to my soundcard. So if I go out and buy loopback cable for as much as $5, then connect audio output to audio input on my full-duplex soundcard and run audio recorder program in parallel with whatever decrypting program they use I still would record the song, encode it into normal MP3 and save it on a Zip disk to listen to it at work. There is no way they could stop me.
Article actually talks about 3 applications with very different approaches:
PacketShaper, etc - software/hardware designed to limit traffic per application based on ports app uses. This is actually a good thing - believe it or not, bandwith costs money ! If I remember correctly, FreeBSD-based firewalls supported bandwith quotas per user for quite lond time. Since Gnutella and at least 2 free/* open-source */ Napster clones allow you to use arbitrary server/port numbers this would eventually boil down to per-user bandwith control.
PacketHound, etc - software/hardware that locates and kills specific protocol. That's evil, of course, and probably freedom to choose protocol is freedom of speech, just like your choice of language/* Do I look like a lawyer ?!? */. However, this approach is not very effective. ssh, for example, allows you to tunnel TCP connections - encrypted, of course, multiple connections over one channel, etc. PacketHound won't puick this up. Actually, adding SSL layer in Gnutella would be great, that would solve this problem once and for all
MediaEnforcer - software athat uses Gnutella/Napster/etc protocols to track which user hosts which files. That software would allow, for example, user alpha@fsb.ru to search for ("Elections results" AND "Faked") and determine that user vpupkin@cityline.ru with ip 123.45.67.89 have material that is not suitable for minors, violates copyright laws and should be deemed as unappropriate as a whole... If you dont know what KGB and it's successor, FSB, do with people who say unappropriate things, consider yourself very lucky ! The problem lies in mechanism used by Gnutella and Napster - they return remote users' IP address instead of routing pieces of information thru network, i.e. if user alpha@fsb.ru sends request to it's closest Gnutella node, russia-relay@eff.org, node returns IP of the machine that has data requested. FreeNet, on the other hand, relays data thru network. For example alpha@fsb.ru requests file elections.html from russia-relay@eff.org, who requests file from vpupkin@cityline.ru and passes it to alpha@fsb.ru. That way every user is not aware about any node other his/her neghbors, and could not determine from which node material originated. That seems to be a comparably good solution, given that encryption is strong enough and protocol does not record any extra information.
huge number of free (open-source) software floating around. You don't even need to borrow any code, you could just see how feature X is implemented (for example I remember looking at some Perl code and implementing smth similar in C)
open standarts. Practically all kinds of Unix are compatible on source level because of that (POSIX, X libs and everything on top of them, i.e. GTK/QT, etc, is the same on every Unix). You don't have huge number of undocumented (or in most cases in Windows half-documented) APIs
Practically all other advantages depends on language you use. Unix have incredibly good set of tools (gcc,make,cvs,gprof,etc). Most of them had been ported to windows as well, I don't have much experience with those ports. However, there are some advantages of Unix system itself:
For scripting languages one big advantage is that kernel automatically runs interpreter specified after magic '#!' . That could be a quite helpful if you have a system with many scripts running each other. For example if 25 programs call script X, and you decided to rewrite it, say, in Phyton istead of Perl, you don't need to modify those 25 programs to run Phyton interpreter - you just modify the script.
For most shell scripts stuff like/proc filesystem is very useful too - you could extract a lot of information from it. Also, windows still lacks many shell utils. Last time I checked, pipes support was horriple - maybe it's changed in W2K, thought
For C/C++ programmer Unix have even more advantages:
Unix memory protection mechanisms - superior to anything I've seen. Most problems I have with C/C++ programs (especially raw C) are memory-related. Unix approach is very flexible: when program tries to access wrong address in memory it receives SIGSEGV or SIGBUS and could either intercept it and handle situation by itself or it could just create core dump (I usually trap SIGSEGV to output values of internal vars at the time of crash). Libraries like Electric Fence take advantage of this mechanism too. Also, under Unix, application couldn't write in system memory. Under windows it's not supposed to - and as you could easily see none of microsoft's absolutely bug-free applications never ever crash windows, right ?
Support for debugging, like core dumps and ptrace(). Core dump allows you to save program state at any point, so that you could open it in debugger later. ptrace() allows you to trace process execution - in windows there are mechanisms to do this but all the debuggers I've seen (from SoftIce to TD) use their own functions, which means windows' default interface is almost useless. Also, you could attach your debugger to already running process, this is very usefull when debugging something like server process (in my case Squid proxy 2.3DEVEL3)
Much better development tools, like gdb and ddd. gdb, for example, is scriptable, you could easily program it to watch when variable changes to a specific value, when it does set temporary breakpoint in a specific function, trace that function until exit, and then if another variable has specific value set another breakpoint. I don't claim to be VC++ guru, but the only way to do it that I know is to press "next" and "step" buttons around 700 times. I tried. VC++ crashed somewhere at 500th.
DDD (Data Display Debugger) is really just a shell on top of gdb (or dbx or perl debugger). It's tread if you have complex data structures in your program and would like to examine them. DDD allows you to run gdb remotely, for example I ssh'ed to server (SunOS SPARC) from my workstation couple of times to debug a running cache proxy.
Some of these things could be implemented under windows, but things like watchpoints would require debugger to rewrite parts of OS.
Tools like gcc and emacs. Emacs Is The Olny One True Editor, period ! Seriously, I never seen any other sools able to do everything emacs can, i.e.:
Syntax hiliting, auto ident, parens balancing, and other basic features is of course a requirement, not an optional feature. Unfortunately, I haven't seen any editor yet that has it all in one place, i.e. expression auto-hiliting, identing, coloring, all the advanced search/replace, etc.
code navigation, i.e. search any identifier in the whole source tree using tags table, speedbar (tree-like browser of functions, structures, etc) bookmark support, etc.
Support for diffs, CVS/RCS, Makefile modes Please note that all the features above are supported at least for C/C++, Perl and Java, function navigation also works with List, etc
Macros: emacs main advantage. To put it simply, emacs have macros system based on Lisp, powerfull enough to write mailer, news reader, web browser, games like tetris, AIM/ICQ/IRC support, telnet/FTP/ssh support, browsing thru compressed archives, etc
I've tried windows version of emacs, but they have very limited number of features compared to Unix versions
As for gcc, it's simply a C/C++/Objective C compiler with support for, if I remember correctly, at least 29 platforms, one of the best optimisers I've ever seen (egcs beats VC++ and Borland on every project I've seen), very powerfull error-detection mechanisms (I always use -Wall -pedantic options, it's way much more picky than either VC++ or Borland CppBuilder, but I never ever seen it bark on something that was not an error) gcc is ported to windows, slow, but it's working
Last, but certainly not the least, user interfaces are better designed. For example, in Unix once you open socket you could use file IO functions like write() to write into it. When I tried to do it on NT I had 2 different results: General Protection Fault and BSOD. Unix system supports some very advanced functions, like for example, locking only certain parts of file (you could lock bytes 30 to 56 in a file, rest of the file is writable). X has very powerful API. One thing I really like is window being able to swallow (capture) another window. That, for instance, allows you to run any X application inside Netscape, provided you have plugin called xswallow. X does more than just windowing, it allows you to create atoms and properties - basically something like windows registry, only transparent thru network. Unix Netscape uses this mechanism to send "remote" commands to it's instance runnings, even if that instance is running on another machine. Similarly, XEmacs uses this interface to to connect to remote emacs process. Unix has wide array of GUI toolkits, like Motif/GTK/QT/Tk. Some of them are much easier (for me) to use than MFC. For example GTK's event handling scheme is much better. GUI development tools that Unix has look much more appealing to me than their windowsa counterparts. For example Glade - tool that creates GUI interfaces visually. There is a library called libglade that allows you to load Glage project on runtime. That way you could have "pluggable" user interfaces. I have yet to see anything like that for windows. Component architecture upon which GNOME is build (CORBA) also seems to be better for me than COM. I'm not a COM expert, but parts of COM design I'm familiar with (for example the "proxy" components needed to marshall request over network) look like a kludge on top of badly designed protocol for me. CORBA is available for windows, but unlike GNOME, which is all based on CORBA, windows is based on COM, so practically you have CORBA and no components for it
Bottom line: learn every OS you can. Code for Unix - more fun;-)
Apache 1.3.4 Configuration Set OPTIM = "-04 -m486" before compiling on 4 x 400 MHz Pentium II Xeon
Samba 2.0.1 Configuration wide links = no That creates a bottleneck in Samba performance, see here
the following processes were running... (kswapd),/sbin/kerneld,syslogd, not sure if that means something, but why they run kerneld with 2.2 kernel ?
On NT side:
Tcpip\Parameters\Tcpwindowsize = 65535 that makes huge boost on network performance, but only on local network where packets don't get lost
Set Logging - "Next Log Time Period" = "When file size reaches 100 MB" Logs on the F: drive (RAID) along with the WebBench data files So basically server does much less logging than Apache - and since it's many small requests, and since Apache writes logs on a non-RAID disk all together it'll be a big bottleneck
Anyone noted anything else wrong with this benchmark ? From all my experience it looks like pure crap
Ok, on my Linux box Netscape 4.5 seems to create files only readable by me (or root, which is sometimes the same) It also seems to delete those forms after some time because I submitted couple of forms (via POST) and all I now have is/tmp/nscomm40-root which is drwxr-xr-x and is an empty dir. It really seems to be windows bug. Besides, if old netscape versions (or libc5 ? never checked that) do create files readable by everyone - couldn't I just write a shell wrapper for netscape which does umask ?
Slashdot Mirror
There's nothing, on the other hand, that stops corporations s.a. Verizon, from monitoring your calls for profit only, uncontrollably.
Secodndly, the government is completely different from this. It exists to advance a political agenda and control every detail of our lives. It has a moral outloook, and if your morals are different you are screwed.
You elect the government. Dont like police beating - vote Harry Browne. Want clean air - vote Ralph Nader. Want to see how dumb a president could be - vote ... oh, never mind that one ...
On the contrary, you do not elect corporations. You are forced to obey whatever they tell you to do pretty much the same way you're forced to obey laws - i.e. techinically you could ignore laws or corporate agreements, but you'll face the consequences
Corporations have a record of non-abuse, and are owned by the people.
Not exactly, but I like your sence of humor. Corporations care about people, help people and protect people. For a list of people browse www.forbes.com.
Opinions are mine only and could change without notice.
However, you could poison the data, i.e. make a lot of invalid entries - like creating new account on Hotmail each time you need to lookup something on MSDN, etc. As signal-to-noise ratio goes down M$ would end up with huge database, most of data in which is outdated and therefore useles for most Evil Purposes(TM).
As for secure trusted email the best I've seen is Lokmail - they support standart PGP.
Opinions are mine only and could change without notice.
But are there only 2 choices ? What if instead of some agency you would have several volunteer-driven groups (sites) targeted at specific audience which would link to information and provide commentary ? And where readers could then discuss it further ? And then you decide which groups you trust to classify information for you ? You pick several sources to filter "censor" information, and submit information you consider valuable.
I believe I heard about site called ... umm Slashdot was it ? Or kuro5hin ?
There you could get "pre-filtered" information that could be then further corrected/discussed in comments. Anyone else around here heard of Slashdot , no ?
Opinions are mine only and could change without notice.
However, if I understand correctly - I can't find any exact description of what .NET internal structure looks like - M$ is using it for procedure calls. .NET does, hopefully I'm wrong.
XML is a right tool for data exchange, CORBA is right for procedure calls, and I thought we're supposed to use right tool for right job, or even beter, have a choice what to use - Java doesn't force you to use CORBA, you could use XML RPC if you want - heck, it you really need you could send Java class as data and let it run on remote machine, with it's own security manager, if you want - sometimes this is the best solution for distributed systems. But Java gives you a choice. I can't find any evidence that
Opinions are mine only and could change without notice.
My guess is sooner or later we'll get binary XML encoding, then some "preparsed SOAP", and then finally something similar to CORBA - or M$ would switch to CORBA, which is unlikely because it means, gods forbid, Java compatibility
Opinions are mine only and could change without notice.
i.e.
Legally speaking I never seen the code, so is this legit ?
Soviet Union is gone, US is the biggest nuclear power on this planet, no enemies to defend from, at least with Army - to defend from all those terrorists that exist according to current Govt., you would need law enforcement, not Army.
So is there any reason for strong military other than justifying budget spending ?
At my current work they gave me machine with NT 4 taking 4 gigs and 9 gigs of empty space. They said I could install anything I want, but they would only support NT and, surprisingly, FreeBSD up to some extent. I decided to go for BSD since I wanted to learn it. So I fixed NT in a usual fassion (mkfs) and had no problems with IT anymore. I also have Linux on laptop, use it in company. I know that almost every tech in my company uses either Linux or BSD, or Solaris for those with Sun boxes, graphic designers use Mac. The only thing they require is for you to be compatible (SAMBA to access printers, have to be able to read .DOC, etc). And officially IT is not supposed to help you if your Linux box is screwed - I know my bos got about 5 different Linux distros from IT when he asked for it - but officially they aren't responsible. Beyong that - who cares what you use as long as you do your work ?
On my previous work, when I was admin myself, I used Linux as universal troubleshooting tool - had it work as print server for Macs and reroute all print jobs to Netware server, etc. Had to fix couple of linux boxes for professors - it was a college work - I know that most other admins never bothered to fix Linux, or for that matter anything that they had any excuse not to fix. That's probably why some IT managers dislike Linux or anything else that is new - they don't know what security measures are needed for system, how to support it, would it interfere with anything (one of my friends used Linux with Netware, named machine/share or smth like that using Russian characters, half of their windows boxes started crashing when trying to browse network).
StegFS allows you to have mandatory number of data layers. w/o knowing password there is no way to determine how many layers are there. They ask for password, you give them one. Or two. Or five. There is no way to determine if you had given them passwords to all layers of data.
M$ doesn't have that kind of UI. That's one thing they definitely should implement - user-customizable UI - before adding any features - be it voice-recognition, eye tracking, or whatever
So, again, this technology is kind of cool, but it's nothing more (or less) than new input method, and it should be up to software developer to decide wether his/her particular application needs it.
P.S. Almost forgot - M$ sucks no matter what !!!!!
Well, this seems to be a problem with any anonymous multiuser information system - attacker could try to disrupt service by posting "noise" data. Since there is no algorithm which could validate wether the file is noise there should be some kind of moderation system. For example Napster/Gnutella/whatever client generating CRC check of file and chatrooms posting correct CRC. Napster should probably implement "preview seek" feature to listen to random part of song - since people usually remember songs by words in a middle, not by intro, anyway. Another feature I really miss in Napster is download resume - because of that there are so many truncated songs - sort of noise, too. Gnutella has this feature. As I said, CRC sums of songs and something like chatrooms with "moderators" providing correct CRC, and people voting would probably cut down noise level. That's similar to what Slashdot does - coments arent's thrown out, but some get higher rating.
Fundamental problem with any digital media protection system is that it must produce decrypted output for the end user.
This means even if Big Brother Records Inc. would require me to use special device attached to computer, that decodes music according to my fingerprints, sound output would still go to my soundcard. So if I go out and buy loopback cable for as much as $5, then connect audio output to audio input on my full-duplex soundcard and run audio recorder program in parallel with whatever decrypting program they use I still would record the song, encode it into normal MP3 and save it on a Zip disk to listen to it at work. There is no way they could stop me.
The problem lies in mechanism used by Gnutella and Napster - they return remote users' IP address instead of routing pieces of information thru network, i.e. if user alpha@fsb.ru sends request to it's closest Gnutella node, russia-relay@eff.org, node returns IP of the machine that has data requested. FreeNet, on the other hand, relays data thru network. For example alpha@fsb.ru requests file elections.html from russia-relay@eff.org, who requests file from vpupkin@cityline.ru and passes it to alpha@fsb.ru. That way every user is not aware about any node other his/her neghbors, and could not determine from which node material originated. That seems to be a comparably good solution, given that encryption is strong enough and protocol does not record any extra information.
Well, if they use Sun tools they of course could
- huge number of free (open-source) software floating around. You don't even need to borrow any code, you could just see how feature X is implemented (for example I remember looking at some Perl code and implementing smth similar in C)
- open standarts. Practically all kinds of Unix are compatible on source level because of that (POSIX, X libs and everything on top of them, i.e. GTK/QT, etc, is the same on every Unix). You don't have huge number of undocumented (or in most cases in Windows half-documented) APIs
- For scripting languages one big advantage is that kernel automatically runs interpreter specified after magic '#!' . That could be a quite helpful if you have a system with many scripts running each other. For example if 25 programs call script X, and you decided to rewrite it, say, in Phyton istead of Perl, you don't need to modify those 25 programs to run Phyton interpreter - you just modify the script.
/proc filesystem is very useful too - you could extract a lot of information from it.
For C/C++ programmer Unix have even more advantages:Practically all other advantages depends on language you use. Unix have incredibly good set of tools (gcc,make,cvs,gprof,etc). Most of them had been ported to windows as well, I don't have much experience with those ports. However, there are some advantages of Unix system itself:
For most shell scripts stuff like
Also, windows still lacks many shell utils. Last time I checked, pipes support was horriple - maybe it's changed in W2K, thought
Unix memory protection mechanisms - superior to anything I've seen. Most problems I have with C/C++ programs (especially raw C) are memory-related. Unix approach is very flexible: when program tries to access wrong address in memory it receives SIGSEGV or SIGBUS and could either intercept it and handle situation by itself or it could just create core dump (I usually trap SIGSEGV to output values of internal vars at the time of crash). Libraries like Electric Fence take advantage of this mechanism too.
Also, under Unix, application couldn't write in system memory. Under windows it's not supposed to - and as you could easily see none of microsoft's absolutely bug-free applications never ever crash windows, right ?
Support for debugging, like core dumps and ptrace(). Core dump allows you to save program state at any point, so that you could open it in debugger later. ptrace() allows you to trace process execution - in windows there are mechanisms to do this but all the debuggers I've seen (from SoftIce to TD) use their own functions, which means windows' default interface is almost useless. Also, you could attach your debugger to already running process, this is very usefull when debugging something like server process (in my case Squid proxy 2.3DEVEL3)
Much better development tools, like gdb and ddd. gdb, for example, is scriptable, you could easily program it to watch when variable changes to a specific value, when it does set temporary breakpoint in a specific function, trace that function until exit, and then if another variable has specific value set another breakpoint. I don't claim to be VC++ guru, but the only way to do it that I know is to press "next" and "step" buttons around 700 times. I tried. VC++ crashed somewhere at 500th.
DDD (Data Display Debugger) is really just a shell on top of gdb (or dbx or perl debugger). It's tread if you have complex data structures in your program and would like to examine them. DDD allows you to run gdb remotely, for example I ssh'ed to server (SunOS SPARC) from my workstation couple of times to debug a running cache proxy.
Some of these things could be implemented under windows, but things like watchpoints would require debugger to rewrite parts of OS.
Tools like gcc and emacs.
Emacs Is The Olny One True Editor, period !
Seriously, I never seen any other sools able to do everything emacs can, i.e.:
Please note that all the features above are supported at least for C/C++, Perl and Java, function navigation also works with List, etc
I've tried windows version of emacs, but they have very limited number of features compared to Unix versions
As for gcc, it's simply a C/C++/Objective C compiler with support for, if I remember correctly, at least 29 platforms, one of the best optimisers I've ever seen (egcs beats VC++ and Borland on every project I've seen), very powerfull error-detection mechanisms (I always use -Wall -pedantic options, it's way much more picky than either VC++ or Borland CppBuilder, but I never ever seen it bark on something that was not an error)
gcc is ported to windows, slow, but it's working
Last, but certainly not the least, user interfaces are better designed. For example, in Unix once you open socket you could use file IO functions like write() to write into it. When I tried to do it on NT I had 2 different results: General Protection Fault and BSOD. Unix system supports some very advanced functions, like for example, locking only certain parts of file (you could lock bytes 30 to 56 in a file, rest of the file is writable). X has very powerful API. One thing I really like is window being able to swallow (capture) another window. That, for instance, allows you to run any X application inside Netscape, provided you have plugin called xswallow. X does more than just windowing, it allows you to create atoms and properties - basically something like windows registry, only transparent thru network. Unix Netscape uses this mechanism to send "remote" commands to it's instance runnings, even if that instance is running on another machine. Similarly, XEmacs uses this interface to to connect to remote emacs process.
Unix has wide array of GUI toolkits, like Motif/GTK/QT/Tk. Some of them are much easier (for me) to use than MFC. For example GTK's event handling scheme is much better.
GUI development tools that Unix has look much more appealing to me than their windowsa counterparts. For example Glade - tool that creates GUI interfaces visually. There is a library called libglade that allows you to load Glage project on runtime. That way you could have "pluggable" user interfaces. I have yet to see anything like that for windows.
Component architecture upon which GNOME is build (CORBA) also seems to be better for me than COM. I'm not a COM expert, but parts of COM design I'm familiar with (for example the "proxy" components needed to marshall request over network) look like a kludge on top of badly designed protocol for me. CORBA is available for windows, but unlike GNOME, which is all based on CORBA, windows is based on COM, so practically you have CORBA and no components for it
Bottom line: learn every OS you can. Code for Unix - more fun ;-)
Set OPTIM = "-04 -m486" before compiling
on 4 x 400 MHz Pentium II Xeon
Samba 2.0.1 Configuration
wide links = no
That creates a bottleneck in Samba performance, see here
the following processes were running ... (kswapd), /sbin/kerneld,syslogd,
not sure if that means something, but why they run kerneld with 2.2 kernel ?
On NT side:
Tcpip\Parameters\Tcpwindowsize = 65535
that makes huge boost on network performance, but only on local network where packets don't get lost
Set Logging - "Next Log Time Period" = "When file size reaches 100 MB"
Logs on the F: drive (RAID) along with the WebBench data files So basically server does much less logging than Apache - and since it's many small requests, and since Apache writes logs on a non-RAID disk all together it'll be a big bottleneck
Anyone noted anything else wrong with this benchmark ?
From all my experience it looks like pure crap
P.S. Why they needed NFS ? inetd ?
Ok, on my Linux box Netscape 4.5 seems to create files /tmp/nscomm40-root which is drwxr-xr-x
only readable by me (or root, which is sometimes the same)
It also seems to delete those forms after some time
because I submitted couple of forms (via POST) and
all I now have is
and is an empty dir.
It really seems to be windows bug. Besides, if
old netscape versions (or libc5 ? never checked that)
do create files readable by everyone - couldn't I
just write a shell wrapper for netscape which does umask ?