Seriously, I haven't had a single spam get through my SpamBayes filter in months, and I rarely get any of my valid mail dumped in "unsure". I see no reason why anyone with a modicum of technical nous should even be caring about spam, unless they are paying for metered bandwidth.
If you saw the first 4, you missed the good stuff
on
The Business of Anime
·
· Score: 1
GITS:Stand Alone Complex does cover exactly the same high-concept material, but it doesn't really pick up steam until several episodes in. The first few are floating stories that are mainly about character establishment.
Is it better? It has more time to delve into characterization and backstory. The arc themes are the same as in the movies, so it's hardly groundbreaking. Depends what you like, I guess.
Anything above writing your own heap allocator over raw OS calls is "a crutch". It's merely a matter of choosing the right crutch.
File handles and network connections? That's none of the GC's business. The algorithm deals with that stuff - GC is there to save you the hassle of asking "who owns this object to delete it" and "is this pointer still good". GC lets you nest function calls in the knowledge intermediate objects won't leak. It lets you grab an arbitrary object and point to it from a data structure without worrying it will become invalid. It's not there to do shutdown housekeeping, that's a C++ism.
There are GC algorithms that can be done in realtime (meaning known bounded time). They're slower, but they exist.
Also don't assume that malloc/free are realtime! malloc/free has to traverse a free list, and will usually coalesce, sort and categorize blocks. new/delete in C++ may also run arbitrary amounts of setup or finalization. Depending how the free list is organized, deallocation order may affect performance in odd ways.
If you're coding for hard realtime embedded apps, you shouldn't malloc. Predefine buffers or design the algorithm so it doesn't use heap.
I can see no sensible reason why firefox is not linked against the Boehm collector. You can use that for hunting leaks, for tidying after leaks, and for skipping the whole manual memory management hassle in the first place. I'm not sure if the Boehm collector does compaction, but a sufficiently clever "conservative" GC can also compact (some) memory.
Nowadays the state of the art has moved on and there is no excuse, none at all, for malloc/free.
Being as insecure as *sh is hardly a selling point! Shell scripts are either traditional and tested, small and heavily debugged, or never put into a position where they can be used for privilege elevation. Nobody sane writes serious user-facing apps in shell. People who put.sh files in their cgi-bin directory deserve to be hacked.
Compare serious scripting languages. Perl supports taint mode, Ruby has $SAFE. They all force you to be explicit when you want to call a an OS command, even if only via backticks.
REXX is worse than all of the above because it pretends to be a serious script language, then stabs you in the back with automatic handoff to OS commands.
Aparrently REXX will call an OS command if it doen't recognise a keyword. So just plain including grep and uniq and so forth as if they were functions is allowed. How quaint! Just think of the hacks, if you misspell pront or writte or opne or some random thing, you could be running any random black-hat code and perhaps never even knowing.
Sorry, nostalgic folks. The state of the art has moved on, and not accidentally.
No, they invented that tech, and after long hard lessons discarded it. As Dr. Non Canonical explains: "[balcony rails] are the first step in a civilizational decline that begins with a laudable concern for safety and ends up with 24 hour nannying of adults, and allergy safety labels reading 'contains nuts' on packets of nuts. The result of cosseting the shallow end of the gene pool is that it breeds and expands, quickly overwhelming literacy, sanity, and self responsibility. Entire planets which travelled this route have been lost, upon the demise from loneliness of the last person who knew how to be a nanny rather than require one. While compassion would suggest we protect those stupid enough to step off their own balcony (or off footbridges, etc), hard practical experience says let them fall." (emphasis in original)
You don't have a right to expect other people privately using a protocol-linked inter-network (public by mutual implied consent) to act in ways that are personally convenient to you. If somebody does you actual harm, or trespasses into your property, go after them. But you have no "freedom" to require others not to do things for their own reasons that coincidentally make harm or trespass easier. That's what's known as "tough luck". Suck it up, or switch off your modem.
Perl had a niche before CGI. It was a general language for throwaway programs and system scripts. You could write simple, small apps and servers in it, by using CPAN to do 90% of the work for you. This also resulted in clean, readable code (the mess being hidden in the modules). My email is downloaded off pop3 and inserted into maildirs by a perl script I designed. It's short and simple, despite using a boatload of tricks to avoid memory leaks and lockups no matter what. It just runs, so I can ignore it.
I find that in its old niche, perl still reigns supreme. Modern deep, complex webapps may have moved beyond its sphere of competence, but I doubt I'll be uninstalling perl anytime soon.
Other people have pointed up the other holes in your theories, so I'll point up something different, namely: you mistakenly assume that only big is legal. Sure the latest Ubuntu liveCD torrent won't shut down from an overloaded tracker. But what about JoeSmithLinux or other hypothetical small distro? Suppose one such is surprisingly good and gets posted to Slashdot. A thundering herd of geeks crams down this guy's DSL and knocks him off the net. Distributed tracker to the rescue!
DT really is just an extension of the BT idea of spreading the load. BT allows downloads to scale, DT allows trackers to scale. It's better for "piracy" simply because it's better for everything. Just as BT was versus HTTP.
While this is an obviuous non-starter, it points up a collision of two trends. First, a limit, the public won't hand over their hard-earned money for an overly intrusive DRM scheme (the original DIVX proved that). Second, the now mature and highly effective P2P distribution infrastructure, which will quickly cut through all non-intrusive DRM.
I don't believe there is a level of DRM, strong enough to work, that the public will tolerate. I don't believe that the *AA will be able to strongarm the market into adopting blu-ray or whatever - they'll just lose so much money trying that they'll have to surrender and release on DVD. I know that politicians, bought or not, don't dare push the public too far.
Sooner or later the only option is going to be: let people copy, because you can't stop them.
What will the *AA do when they realize their bind?
What have you done to deserve this? Stuck with Slashdot, and put up with its faults. You want a better "REAL news for nerds, stuff that ACTUALLY matters", go start one.
No, seriously. If you think/. is getting a free ride, time to set up in competition.
...that project Harmony was the reason TrollTech chose to GPL (as versus seeing their strategic role usurped by an LGPL workalike). At which point Harmony dried up as redundant. So while it didn't per se do much, its historic impact isn't negligible.
M$.doc format is used by businesses sending documents around, but it's not a particularly good tool for the task. Business makes use of it because they pretty much have to, but nobody in their right mind would deliberately transform documents into.doc and back out again. The chances of losing information are just too high, even between versions of MS Word.
Compare Open Document. The specification is stable and transparent. Use of ODF as an intermediate format in workflows looks eminently sensible. When you need it in another format, it can be rendered into that format at the last possible moment. Converters need only be written once and each new converter adds another potential input/output to anybody's workflow. Tools can be written to understand ODF, pick it apart and do stuff. Tools can be strung end-to-end in pipelines, "unix philosophy" style.
In other words it could become to live documents what PDF is to "digital printout", and more.
M$ will support it as they are gradually pushed out onto the margins. Yes Joe Cubicle will still use.doc files, but the real money will be in the steps before and after he gets the file, and that will all be ODF.
I've always thought of BT's lack of built-in search as an asset. Distributed search is dreck, or at least all modern incarnations thereof. The problems of file quality, mesh instability, bandwidth waste, info reachability cannot be solved, because they're inherent in working with a sparsely-connected crowd of strangers.
Distributed search seems to me to be an artefact of its time. Then the choices were a single vulnerable center, or compromise utility to remove the center. Napster versus Gnutella.
More modern BT has many centers with seperation of function so files can be seeded from one box, tracked on another, and listed on a third. BT's search sites like suprnova are centralized (with all the advantages), but expendable. The sum of all search sites is effectively immortal.
We are a species that, naked, dies from cold or sun. Our clothes are our skin. So are contact lenses our eyes, and cars our feet?
If anything this and similar are gradually working to deconstruct the idea of the natural born human as a standalone unit. Rather, all humans are necessarily "cyborgs". Creating and integrating with tools to extend the self is the true species specialty.
Seriously, if you want to know why Azureus is so damn cool, just click on the new "swarm" tab on a running torrent. Ye gods, but that's beauty! Perfectly abstract, instantly comprehensible, informative in realtime, mesmerizing as a screensaver. You have to respect the kind of people who'd think up something like that.
(Karma bonus turned off because this is OT, but damn, I just had to say that.)
I don't know the details, but I can reasonably assume this new system does the same thing, maintains a partially connected mesh. Otherwise on really popular torrents you might spend the first half hour downloading IPs! That would just be silly.
1) If a big, high-bandwidth tracker dies, everybody can keep downloading, and can even create a URL that lets new people join the torrent.
2) If a little guy without the bandwidth to run a mega-tracker wants to create a torrent, and for whatever reason doesn't want to upload to a public tracker, he can publish a distributed URL which shares the burden of both downloads and tracking amongst the other peers. I imagine this will be particularly useful in limited-interest torrents. For example in intranets, or if the torrent is such that it would be booted from a public tracker (for being eg: too niche).
3) It simplifies the job of torrent-aggregating sites too. They don't need to publish torrent files or maintain trackers. They can just list these textual URLs and let Azureus do the heavy lifting.
They won't be culling the humanlike critters. Only the ones still naive enough to show it, after seeing their buddies getting the ol' hypodermic o' doom.
In other words, they'd be forcing the evolution of smart sneaky mice. And won't that be a good thing, boys and girls?
J. Morrison's corollary to Greenspun's tenth rule
on
Practical Common Lisp
·
· Score: 1
All real-world Common Lisp implementations are bug-ridden, slow and generally implement about half of Common Lisp, plus sufficient ad-hoc informally-specified extensions to drag it into the modern era.
Slashdot Mirror
Seriously, I haven't had a single spam get through my SpamBayes filter in months, and I rarely get any of my valid mail dumped in "unsure". I see no reason why anyone with a modicum of technical nous should even be caring about spam, unless they are paying for metered bandwidth.
GITS:Stand Alone Complex does cover exactly the same high-concept material, but it doesn't really pick up steam until several episodes in. The first few are floating stories that are mainly about character establishment.
Is it better? It has more time to delve into characterization and backstory. The arc themes are the same as in the movies, so it's hardly groundbreaking. Depends what you like, I guess.
...collect it all into a convenient library, for the use of sneaky b*stards.
I propose the name "libgood.so"
See here.
Anything above writing your own heap allocator over raw OS calls is "a crutch". It's merely a matter of choosing the right crutch.
File handles and network connections? That's none of the GC's business. The algorithm deals with that stuff - GC is there to save you the hassle of asking "who owns this object to delete it" and "is this pointer still good". GC lets you nest function calls in the knowledge intermediate objects won't leak. It lets you grab an arbitrary object and point to it from a data structure without worrying it will become invalid. It's not there to do shutdown housekeeping, that's a C++ism.
There are GC algorithms that can be done in realtime (meaning known bounded time). They're slower, but they exist.
Also don't assume that malloc/free are realtime! malloc/free has to traverse a free list, and will usually coalesce, sort and categorize blocks. new/delete in C++ may also run arbitrary amounts of setup or finalization. Depending how the free list is organized, deallocation order may affect performance in odd ways.
If you're coding for hard realtime embedded apps, you shouldn't malloc. Predefine buffers or design the algorithm so it doesn't use heap.
I can see no sensible reason why firefox is not linked against the Boehm collector. You can use that for hunting leaks, for tidying after leaks, and for skipping the whole manual memory management hassle in the first place. I'm not sure if the Boehm collector does compaction, but a sufficiently clever "conservative" GC can also compact (some) memory.
Nowadays the state of the art has moved on and there is no excuse, none at all, for malloc/free.
Being as insecure as *sh is hardly a selling point! Shell scripts are either traditional and tested, small and heavily debugged, or never put into a position where they can be used for privilege elevation. Nobody sane writes serious user-facing apps in shell. People who put .sh files in their cgi-bin directory deserve to be hacked.
Compare serious scripting languages. Perl supports taint mode, Ruby has $SAFE. They all force you to be explicit when you want to call a an OS command, even if only via backticks.
REXX is worse than all of the above because it pretends to be a serious script language, then stabs you in the back with automatic handoff to OS commands.
Aparrently REXX will call an OS command if it doen't recognise a keyword. So just plain including grep and uniq and so forth as if they were functions is allowed. How quaint! Just think of the hacks, if you misspell pront or writte or opne or some random thing, you could be running any random black-hat code and perhaps never even knowing.
Sorry, nostalgic folks. The state of the art has moved on, and not accidentally.
...and get it done to you for free (if you're not picky).
No, they invented that tech, and after long hard lessons discarded it. As Dr. Non Canonical explains: "[balcony rails] are the first step in a civilizational decline that begins with a laudable concern for safety and ends up with 24 hour nannying of adults, and allergy safety labels reading 'contains nuts' on packets of nuts. The result of cosseting the shallow end of the gene pool is that it breeds and expands, quickly overwhelming literacy, sanity, and self responsibility. Entire planets which travelled this route have been lost, upon the demise from loneliness of the last person who knew how to be a nanny rather than require one. While compassion would suggest we protect those stupid enough to step off their own balcony (or off footbridges, etc), hard practical experience says let them fall." (emphasis in original)
You don't have a right to expect other people privately using a protocol-linked inter-network (public by mutual implied consent) to act in ways that are personally convenient to you. If somebody does you actual harm, or trespasses into your property, go after them. But you have no "freedom" to require others not to do things for their own reasons that coincidentally make harm or trespass easier. That's what's known as "tough luck". Suck it up, or switch off your modem.
Perl had a niche before CGI. It was a general language for throwaway programs and system scripts. You could write simple, small apps and servers in it, by using CPAN to do 90% of the work for you. This also resulted in clean, readable code (the mess being hidden in the modules). My email is downloaded off pop3 and inserted into maildirs by a perl script I designed. It's short and simple, despite using a boatload of tricks to avoid memory leaks and lockups no matter what. It just runs, so I can ignore it.
I find that in its old niche, perl still reigns supreme. Modern deep, complex webapps may have moved beyond its sphere of competence, but I doubt I'll be uninstalling perl anytime soon.
Other people have pointed up the other holes in your theories, so I'll point up something different, namely: you mistakenly assume that only big is legal. Sure the latest Ubuntu liveCD torrent won't shut down from an overloaded tracker. But what about JoeSmithLinux or other hypothetical small distro? Suppose one such is surprisingly good and gets posted to Slashdot. A thundering herd of geeks crams down this guy's DSL and knocks him off the net. Distributed tracker to the rescue!
DT really is just an extension of the BT idea of spreading the load. BT allows downloads to scale, DT allows trackers to scale. It's better for "piracy" simply because it's better for everything. Just as BT was versus HTTP.
While this is an obviuous non-starter, it points up a collision of two trends. First, a limit, the public won't hand over their hard-earned money for an overly intrusive DRM scheme (the original DIVX proved that). Second, the now mature and highly effective P2P distribution infrastructure, which will quickly cut through all non-intrusive DRM.
I don't believe there is a level of DRM, strong enough to work, that the public will tolerate. I don't believe that the *AA will be able to strongarm the market into adopting blu-ray or whatever - they'll just lose so much money trying that they'll have to surrender and release on DVD. I know that politicians, bought or not, don't dare push the public too far.
Sooner or later the only option is going to be: let people copy, because you can't stop them.
What will the *AA do when they realize their bind?
If you misspell the error messages differently in each possible error source, then grep will find your bug instantly :-)
What have you done to deserve this? Stuck with Slashdot, and put up with its faults. You want a better "REAL news for nerds, stuff that ACTUALLY matters", go start one.
/. is getting a free ride, time to set up in competition.
No, seriously. If you think
...that project Harmony was the reason TrollTech chose to GPL (as versus seeing their strategic role usurped by an LGPL workalike). At which point Harmony dried up as redundant. So while it didn't per se do much, its historic impact isn't negligible.
M$ .doc format is used by businesses sending documents around, but it's not a particularly good tool for the task. Business makes use of it because they pretty much have to, but nobody in their right mind would deliberately transform documents into .doc and back out again. The chances of losing information are just too high, even between versions of MS Word.
.doc files, but the real money will be in the steps before and after he gets the file, and that will all be ODF.
Compare Open Document. The specification is stable and transparent. Use of ODF as an intermediate format in workflows looks eminently sensible. When you need it in another format, it can be rendered into that format at the last possible moment. Converters need only be written once and each new converter adds another potential input/output to anybody's workflow. Tools can be written to understand ODF, pick it apart and do stuff. Tools can be strung end-to-end in pipelines, "unix philosophy" style.
In other words it could become to live documents what PDF is to "digital printout", and more.
M$ will support it as they are gradually pushed out onto the margins. Yes Joe Cubicle will still use
I've always thought of BT's lack of built-in search as an asset. Distributed search is dreck, or at least all modern incarnations thereof. The problems of file quality, mesh instability, bandwidth waste, info reachability cannot be solved, because they're inherent in working with a sparsely-connected crowd of strangers.
Distributed search seems to me to be an artefact of its time. Then the choices were a single vulnerable center, or compromise utility to remove the center. Napster versus Gnutella.
More modern BT has many centers with seperation of function so files can be seeded from one box, tracked on another, and listed on a third. BT's search sites like suprnova are centralized (with all the advantages), but expendable. The sum of all search sites is effectively immortal.
We are a species that, naked, dies from cold or sun. Our clothes are our skin. So are contact lenses our eyes, and cars our feet?
If anything this and similar are gradually working to deconstruct the idea of the natural born human as a standalone unit. Rather, all humans are necessarily "cyborgs". Creating and integrating with tools to extend the self is the true species specialty.
Seriously, if you want to know why Azureus is so damn cool, just click on the new "swarm" tab on a running torrent. Ye gods, but that's beauty! Perfectly abstract, instantly comprehensible, informative in realtime, mesmerizing as a screensaver. You have to respect the kind of people who'd think up something like that.
(Karma bonus turned off because this is OT, but damn, I just had to say that.)
I don't know the details, but I can reasonably assume this new system does the same thing, maintains a partially connected mesh. Otherwise on really popular torrents you might spend the first half hour downloading IPs! That would just be silly.
...that's still a valid question.
Some answers:
1) If a big, high-bandwidth tracker dies, everybody can keep downloading, and can even create a URL that lets new people join the torrent.
2) If a little guy without the bandwidth to run a mega-tracker wants to create a torrent, and for whatever reason doesn't want to upload to a public tracker, he can publish a distributed URL which shares the burden of both downloads and tracking amongst the other peers. I imagine this will be particularly useful in limited-interest torrents. For example in intranets, or if the torrent is such that it would be booted from a public tracker (for being eg: too niche).
3) It simplifies the job of torrent-aggregating sites too. They don't need to publish torrent files or maintain trackers. They can just list these textual URLs and let Azureus do the heavy lifting.
They won't be culling the humanlike critters. Only the ones still naive enough to show it, after seeing their buddies getting the ol' hypodermic o' doom.
In other words, they'd be forcing the evolution of smart sneaky mice. And won't that be a good thing, boys and girls?
All real-world Common Lisp implementations are bug-ridden, slow and generally implement about half of Common Lisp, plus sufficient ad-hoc informally-specified extensions to drag it into the modern era.