I actually wrote my own book as sort of a prequel to TAOCP. TAOCP assumes that you already have experience with some sort of assembly language. My book doesn't assume anything:
"Also, to nitpick a bit on his point, to be able to replay the bits, you first do have to record them, which equates to a man-in-the-middle attack. This should be able to be avoided by some simple public/private cryptography built in to the device."
Not really. You will _always_ have a stage where the stuff is not encoded. If you can get my bioinformatic data once, I'm totally screwed, because I can't change my password to something else. My security will be forever broken.
Think about the current issues with ATM cards. People put in their own devices on top of the ATM machines, and just read the contents of the ATM card. In fact, with bioinformatics, I don't even need to get that close, because your eyes and fingerprints are on everything. All I need to do is be able to shortcut the reader and I'm all set. The security moves from being in the software (which is often remote) to being in the hardware, which is local.
The problem w/ biometrics is that it will wind up being way too easy to bypass (by just recording someone else's bits and replaying them to the hardware, or it will require too much money to secure the biometrics device.
I had heard of a password mechanism once that was based on facial recognition which seemed interesting. You chose a sequence of faces, and the computer asks you to choose a face from a selection. It sounded interesting. If anyone knows where the article is, I'd like to re-read on that topic.
" Where's the equal time for creationism? I don't believe in this "evolution" stuff."
The funny thing is that almost any time when the word "evolution" is used in modern day lingo apart from biology, "creation" is in fact what they meant.
My favorite is Linus Torvalds saying Linux is the result of evolution. His thinking is that since the pressures on Linux developers are untamed, it is like evolution, but in fact, evolution is the opposite. In evolution, it is the development which is random, and the pressures which are fairly steady.
This is often known as "Berra's Blunder" after biologist Tim Berra, who used similarities in Corvettes as proof of evolution. It seems that Berra was unaware that Corvettes were designed, and their similarities were based on similarity of designer and requirements, not of heredity:)
What I didn't like about the PDF one (haven't looked at the other) is that Python is missing a lot of influences. It has influences from both Lisp and Smalltalk, and considers itself to be the modern-day version of each.
When Dillo grows up to be a real browser (like, maybe, implements CSS or something [it does a little but not much]), then maybe it will be worth it to worry about supporting it.
And besides, even if you used Dillo for viewing the web, you could still set up Mozilla as your external player for Flash videos:)
"Flash implies another support requirement, something that this guy is not interested in, specifically."
Not any more than a media player. A media player is much more of a support requirement, because I would guess that more people have flash installed than media players. In addition, he wouldn't need to support a different media player for each platform, just flash for all of them.
I've found MPEG1 to be the most compatible across all players.
However, there is another codec that you play in flash instead of in a video player - the flash video format. Macromedia Flash comes with a video compressor to automatically generate a flash video which can be trivially inserted into a flash movie.
The fact that such language is used does not mean that they don't expect everyone else to follow along. Just look at how much influence weather-based science, which is still in its infancy, has on public policy, when it really should have very little influence at this stage in our knowledge.
"Scientist change their minds because that's the way science works. You get new information and experiments and you attempt to understand it in the framework of current theories."
I think the problem that some of us have is that scientists expect the rest of us to take what they say as fact, when in fact, it is just the best guess they can make today. All of science is suffering because the force of fact of experimental science is being used in the non-experimental sciences. Descriptive, historical, and other sciences are good, but they should not carry the same weight as experimental science.
It is pure guesswork that one even could conceivably model a global weather system for any useful timespan. The fact that we are expected to take these models as fact is just idiocy. I respect the fact that scientists should and need to try out theories and revise them, but I think they should also respect the rest of us to not take them seriously at this stage in their development.
"Maybe the solution is to write a special version of C compiler with built in support for virtual functions in structs, so that people can use a clean, concise syntax for calling them and structs written by different people interoperate naturally. Oh wait..."
They called it Objective C. A similar language arose but which threw out a lot of the low-level C stuff and they called it Java. Then someone came along and puked up their guts and they called the resulting mess C++.
Of course, some of us were using Scheme and just ignored everyone else:)
"In previous versions of PostgreSQL, the pg_dump and pg_restore tools were not very good - dumps that included tables or views often would fail on reimport because PostgreSQL wouldn't know the order in which to import everything."
I haven't had that problem w/ 7.4
"You also had to pass in a number of options on the command line just to get a dump that made sense"
Really? pg_dump and pg_dumpall work just fine for me. As long as you use bytea's instead of the other large object types, that is.
"and large object support was kind of clunky."
Depends. If you use bytea columns (up to 2GB I think) then its super-simple. They have large objects themselves, and they _are_ terrible, but you can generally just use bytea's instead.
" I agree that PostgreSQL sometimes takes itself out for a spin--the optimizer seems to get more confused by some queries in PostgreSQL than in other databases."
Sometimes this can happen if your queries do not include enough type information. 8.0 fixes this, though.
For example, if I have int8 columns, you have to cast any direct numbers into int8's or the optimizer won't find the indexes. Why? Because numbers by default are int's, and therefore can't use int8 indexes. PG8 is smarter about this, but you could work around it by doing 23424243::int8. Note that you don't have to do this with columns comparisons, since the type information for both columns is already there -- only when you are comparing a column of type int8 to a constant value, because they are interpretted as being int4's.
"ut for mission-critical enterprise apps that need 100% uptime (4 nines anyway), you have to have active-active databases, backups while the app is running, etc."
I used to work at EDS back around 2001 and at that time Oracle clustering sucked majorly. It often did not perform as advertised.
Also, I'm pretty sure with this version you can do backups while the app is running (NOTE - you could always do them w/ dumps, but I think now you can even do them w/ binaries + transaction logs).
That's the point you're missing. "object relational" DOES NOT mean "object oriented". They are different. PostgreSQL is object relational while other systems may be object oriented. SQL access does not make a system relational, either, although it does in many people's minds.
You know, many of us are using computers between 200 and 400 MHz. "As slow as 800" should read "as fast as 800" for many of us, especially if we're cheap.
"Except that a stock option is not really a "cost"; it does not deplete the company's assets to issue them."
Two things: (1) you cannot give infinite options, so it is indeed somewhat like a cost, and (2) you are giving options in exchange for work. Without a valuation on the options, you don't have an accurate measure what the cost of your employees are. If I had a highly increasing stock, and I had employees that would just work for options, could I honestly say that my labor cost was zero? If my stock didn't continue to go up, I'd have to pay them the equivalent of what they valued their options as. Therefore, if someone is reading your financial statements, they wouldn't know the ongoing cost of labor, and assume it to be much smaller than it really is.
Slashdot Mirror
If you're going to post the free URL, you should at least post the official one:
http://savannah.nongnu.org/projects/pgubook/
I actually wrote my own book as sort of a prequel to TAOCP. TAOCP assumes that you already have experience with some sort of assembly language. My book doesn't assume anything:
My book teaches programming using assembly language on Linux
"Also, to nitpick a bit on his point, to be able to replay the bits, you first do have to record them, which equates to a man-in-the-middle attack. This should be able to be avoided by some simple public/private cryptography built in to the device."
Not really. You will _always_ have a stage where the stuff is not encoded. If you can get my bioinformatic data once, I'm totally screwed, because I can't change my password to something else. My security will be forever broken.
Think about the current issues with ATM cards. People put in their own devices on top of the ATM machines, and just read the contents of the ATM card. In fact, with bioinformatics, I don't even need to get that close, because your eyes and fingerprints are on everything. All I need to do is be able to shortcut the reader and I'm all set. The security moves from being in the software (which is often remote) to being in the hardware, which is local.
The problem w/ biometrics is that it will wind up being way too easy to bypass (by just recording someone else's bits and replaying them to the hardware, or it will require too much money to secure the biometrics device.
I had heard of a password mechanism once that was based on facial recognition which seemed interesting. You chose a sequence of faces, and the computer asks you to choose a face from a selection. It sounded interesting. If anyone knows where the article is, I'd like to re-read on that topic.
" Where's the equal time for creationism? I don't believe in this "evolution" stuff."
:)
The funny thing is that almost any time when the word "evolution" is used in modern day lingo apart from biology, "creation" is in fact what they meant.
My favorite is Linus Torvalds saying Linux is the result of evolution. His thinking is that since the pressures on Linux developers are untamed, it is like evolution, but in fact, evolution is the opposite. In evolution, it is the development which is random, and the pressures which are fairly steady.
This is often known as "Berra's Blunder" after biologist Tim Berra, who used similarities in Corvettes as proof of evolution. It seems that Berra was unaware that Corvettes were designed, and their similarities were based on similarity of designer and requirements, not of heredity
The Smalltalk implementation Squeak is AWESOME. With their eToys you can do amazing things without typing any code. It is truly a wonder to behold.
What I didn't like about the PDF one (haven't looked at the other) is that Python is missing a lot of influences. It has influences from both Lisp and Smalltalk, and considers itself to be the modern-day version of each.
"As they say, if you're going to bother changing OS's, go whole hog. :-)"
I'm pretty sure I've never heard "them" say that, for any definition of "they".
Jon
Get an iBook. $1,000 will get you a pretty nice machine w/ WiFi, CDRW, 256MB RAM, a 30gig hard drive and OSX.
When Dillo grows up to be a real browser (like, maybe, implements CSS or something [it does a little but not much]), then maybe it will be worth it to worry about supporting it.
:)
And besides, even if you used Dillo for viewing the web, you could still set up Mozilla as your external player for Flash videos
"Flash implies another support requirement, something that this guy is not interested in, specifically."
Not any more than a media player. A media player is much more of a support requirement, because I would guess that more people have flash installed than media players. In addition, he wouldn't need to support a different media player for each platform, just flash for all of them.
Don't know about batch automation, but it took our guys about an hour or two to build playback controls that we could reuse.
Almost all browsers support flash, and flash has its own video format which is pretty nice.
I've found MPEG1 to be the most compatible across all players.
However, there is another codec that you play in flash instead of in a video player - the flash video format. Macromedia Flash comes with a video compressor to automatically generate a flash video which can be trivially inserted into a flash movie.
The fact that such language is used does not mean that they don't expect everyone else to follow along. Just look at how much influence weather-based science, which is still in its infancy, has on public policy, when it really should have very little influence at this stage in our knowledge.
"Scientist change their minds because that's the way science works. You get new information and experiments and you attempt to understand it in the framework of current theories."
I think the problem that some of us have is that scientists expect the rest of us to take what they say as fact, when in fact, it is just the best guess they can make today. All of science is suffering because the force of fact of experimental science is being used in the non-experimental sciences. Descriptive, historical, and other sciences are good, but they should not carry the same weight as experimental science.
It is pure guesswork that one even could conceivably model a global weather system for any useful timespan. The fact that we are expected to take these models as fact is just idiocy. I respect the fact that scientists should and need to try out theories and revise them, but I think they should also respect the rest of us to not take them seriously at this stage in their development.
XML is actually a great idea for structured documents. That's about it, but it's great for that. Combined w/ DSSSL is pure sweetness.
Lisp Macros. Also, Lisp has eval-when, which is pretty cool. Perhaps not quite what Forth has, but still pretty killer.
"Maybe the solution is to write a special version of C compiler with built in support for virtual functions in structs, so that people can use a clean, concise syntax for calling them and structs written by different people interoperate naturally. Oh wait..."
:)
They called it Objective C. A similar language arose but which threw out a lot of the low-level C stuff and they called it Java. Then someone came along and puked up their guts and they called the resulting mess C++.
Of course, some of us were using Scheme and just ignored everyone else
"In previous versions of PostgreSQL, the pg_dump and pg_restore tools were not very good - dumps that included tables or views often would fail on reimport because PostgreSQL wouldn't know the order in which to import everything."
I haven't had that problem w/ 7.4
"You also had to pass in a number of options on the command line just to get a dump that made sense"
Really? pg_dump and pg_dumpall work just fine for me. As long as you use bytea's instead of the other large object types, that is.
"and large object support was kind of clunky."
Depends. If you use bytea columns (up to 2GB I think) then its super-simple. They have large objects themselves, and they _are_ terrible, but you can generally just use bytea's instead.
" I agree that PostgreSQL sometimes takes itself out for a spin--the optimizer seems to get more confused by some queries in PostgreSQL than in other databases."
Sometimes this can happen if your queries do not include enough type information. 8.0 fixes this, though.
For example, if I have int8 columns, you have to cast any direct numbers into int8's or the optimizer won't find the indexes. Why? Because numbers by default are int's, and therefore can't use int8 indexes. PG8 is smarter about this, but you could work around it by doing 23424243::int8. Note that you don't have to do this with columns comparisons, since the type information for both columns is already there -- only when you are comparing a column of type int8 to a constant value, because they are interpretted as being int4's.
"ut for mission-critical enterprise apps that need 100% uptime (4 nines anyway), you have to have active-active databases, backups while the app is running, etc."
I used to work at EDS back around 2001 and at that time Oracle clustering sucked majorly. It often did not perform as advertised.
Also, I'm pretty sure with this version you can do backups while the app is running (NOTE - you could always do them w/ dumps, but I think now you can even do them w/ binaries + transaction logs).
"That's not really all that object relational"
"To see a true object oriented DB..."
That's the point you're missing. "object relational" DOES NOT mean "object oriented". They are different. PostgreSQL is object relational while other systems may be object oriented. SQL access does not make a system relational, either, although it does in many people's minds.
"that's systems as slow as 800mhz."
You know, many of us are using computers between 200 and 400 MHz. "As slow as 800" should read "as fast as 800" for many of us, especially if we're cheap.
"Except that a stock option is not really a "cost"; it does not deplete the company's assets to issue them."
Two things: (1) you cannot give infinite options, so it is indeed somewhat like a cost, and (2) you are giving options in exchange for work. Without a valuation on the options, you don't have an accurate measure what the cost of your employees are. If I had a highly increasing stock, and I had employees that would just work for options, could I honestly say that my labor cost was zero? If my stock didn't continue to go up, I'd have to pay them the equivalent of what they valued their options as. Therefore, if someone is reading your financial statements, they wouldn't know the ongoing cost of labor, and assume it to be much smaller than it really is.