I try to keep backups of data on another machine at my house or in another location entirely.
I totally agree with this. I was always a bit uncomfortable leaving my house until I got all of my automated backup systems in place. Nightly DB dumps and mail tars are encrypted in chunks and rsynced off site. Every night. Every commit to my arch repository is on two machines before I get my prompt back, in another city in about four hours, and another continent about two hours after that. One time setup, and I don't lose data from theft or damage.
Take DHCP for example - damn handy system, developed by microsoft.
Whence did this misinformation originate? I had a similar statement from some MS weenie at my last company. I showed him every bit of protocol documentation I could find at the time and asked him to show me the word ``Microsoft'' in any of it. The closest I found was a windows-specific extension somewhere.
DHCP is an extenion of bootp. They didn't do that, either (see RFC 951, 1534, 1542, 2131 etc...)
It still amazes me that there still aren't any really killer IMAP-implementations around yet. They all seem to suffer from different problems making them annoying to use in the long run.
I know what you mean. I've tried many things. I always went back to pine. Mail.app has been Good Enough(tm) to keep me away from pine most of the time (not that I've got anything against pine).
IMAP's pretty amazing, though. I'd hate to try to implement it again.:)
Java isn't magic here. There's also ocaml, haskell, erlang, python, and countless other languages that prevent fatal accidents while still being very pleasant to use. OCaml, in particular, would probably yield a faster, safer CVS.
Or if nothing else move some of the mail to a backup directory so the poor little imap server doesn't have to deal with YOUR pack-rat habits!
Heh, actually, I have been considering that. Mail.app does need to look at all of the folders every time it connects to see if anything's changed. It's working pretty well right now, but I am considering making a new mail server. Has anything good come out since SunOS 4.1.4?:)
Seriously, though, keeping my mail has come in way too handy for me to consider throwing it away.
Sorry, but I couldn't let this one slide. You've obviously got a special interpretation of "fast", because I tried migrating my Eudora mailboxes to Mail, on a 1Ghz Powerbook G4.
Mail CHOKED on them.
Everyone's got a story and a counter-story. I've got over 100,000 messages in IMAP (101,269 as of last night, but it goes up and down), fully synced to Mail.app (bodies and attachments) indexed for searching, and used every day. It's split over 250 mail boxes (one for each month I've sent or received email as long as I've been keeping stuff).
It's amazingly fast. It makes my mail server seem fast (Sun IPX running SunOS 4.1.4 with a custom cyrus IMAPd that supports compressed mail stores and LDAP and some other stuff).
Imagine if everyone settled on FireWire for the high-speed peripheral bus, and USB never got a chance? We wouldn't have the benifits of USB, namely bus-powered devices, lower cost, support for many devices on the same bus; and then much later, high-speed USB which can finally compete with FireWire regarding bandwidth.
This makes no sense. One of the benefits of USB is bus-powered devices? Like my iPod?
Lower cost? What makes USB lower cost than firewire (cost != price)?
Many devices on the same bus? Like my video camera being controlled by my powerbook as it spools video off onto an external disk (or two)?
High-speed USB that's theoretically similar in speed to firewire being developed while the new firewire standards were being developed is a benefit? That makes the latest USB (theoretically) a little more than half the speed of the latest firewire.
I mean, I'm all for competition and stuff, but USB never seemed to be in the same space as firewire.
With arch, for example, I've branched projects on my laptop (with my standard replicates to a webdav server and a plain http server), made whatever changes I want in a few commits, tracked more changes from head-of-line, and eventually got my changes brought into the original project (doing a merge from my tree).
I.e. anyone, anywhere can branch the tree for any reason and track other people's changes. Additionally, you can track changes from more than one tree at a time. For example, you can track head-of-line, some guy's branch that has feature you want, another guy's branch with another feature, and a third guy's branch with a bug fix. More importantly, it's easy enough that people are actually likely to try that.
Try -O2. I believe O3 includes some optimizations which generally make things run faster, but not always. Aggressive optimization is a chancy thing.
I did. If you can find a configuration that has the C version running as fast as the java version, post it (and explain it if you can...but grabbing an old version of java doesn't count).
Here's what I got (of course, pay no attention to the real time, I was doing other stuff with the machine at the time):
To use this recursive fibonacci-algorithm to compare functional languages and imperative languages is to cheat.
I don't *believe* it works the way you described (that'd make it many times faster, not just a few percent faster (23.07u vs. 31.79u)), but the point is that dealing with every possible detail only makes it harder on the programmer and harder on the compiler to come up with a better way of doing the work.
I agree that fibonacci is a dumb comparison, but it is a top-of-my-head case where you lose a lot in C (i.e. the ``fast'' language) vs. other languages.
The other thing to remember is that Java isn't a functional language.
OCaml will compile to its own bytecode or native code (i.e. not via C).
but what JVM?
java version "1.4.2_03" Java(TM) 2 Runtime Environment, Standard Edition (build 1.4.2_03-117.1) Java HotSpot(TM) Client VM (build 1.4.2-34, mixed mode)
(OS X)
What Compiler was the C code compiled with (with what optimization options)?
gcc version 3.3 20030304 (Apple Computer, Inc. build 1640)
With -O3
Did you try other compilers/JVMs?
I tried it after reading someone's post on orkut suggesting that the java one was faster than the gcc one. The original poster was, I believe, using Linux. I had the same results on my Mac.
Was the JVM already resident in memory on your system?
I did multiple runs of each, so stuff was probably cached.
Because people sure aren't learning them from Java. C hackers are getting harder to find.
Good riddance. I'm getting annoyed by this constant assertion that C is the only real programming. It's just inappropriate for most of what people are doing with it.
For any chunk of C code using malloc() and free(), there's a smaller and faster OCaml equivalent that's garbage collected.
For fun, go to the Great Computer Language shootout and download the java, c, and ocaml fibonacci code (same algorithm). On my machine, the ocaml version is the fastest, followed by the java version, and then the C version. Including startup time, the java version is always faster (user and real CPU time) than the C version.
I write enough C, but OCaml isn't the first language that's produced faster code than I've been able to produce with a C compiler, nor is it the first that's made it easier to write reliable apps than unreliable apps.
What I really don't understand is the way people refer to free() and malloc() as if they're some sort of fundamental elements of programming that are required to be taken seriously. You know they're library routines, right? Should I not take someone seriously who doesn't use sbrk() directly in their code?
Now they have a collaboration of some description with Micro$oft; it's hard to get an ally with more punch than them
Has anyone partnered with MS in any way and come out ahead? I've certainly watched plenty of MS deals go really, really wrong. Three at my last company. The best case, it just cost us money on development. Including things like modifying our web servers so their response headers didn't indicate they were Apache because Microsoft wouldn't link to a web site that wasn't running their own software.
We lost millions and our chance at a business in the worst cases. They treated us as a pilot on our development dime with the promise of an exclusive sales channel or something. When the business proved to be succesful, they made their own. The ones that failed, they just ditched.
Yet companies turn into this reverse Robin Hood thing. Most companies, it seems, will gladly go broke giving money to MS.
I'm sitting at the X Developers' Conference right now watching a presentation by some of Sun's guys on Project Looking Glass and I have to admit that this is some pretty cool shit.
I agree, but having read the Croquet documentation from a few years ago, I wouldn't exactly call it innovative.
I mean, the documentation includes screen shots of Alice looking at a mirror showing other Wonderland creatures.
I have to think Looking Glass was, um, ``inspired'' by the squeak/croquet work.
'd hate to see how you flip out when you get junk mail in real life! Do you call up these companies and threaten to sue?
That's a bad analogy. I don't get hundreds of offensive articles of junk mail every day for each person in my house. If I did, I'd do something about it.
When I go on vacation and try to check my email, it really shows me how much time it wastes. I have server-side spam filters and client-side spam filters. The server-side stuff gets a lot of the spam out of the way, but I was sitting in a hotel a couple of days ago waiting for my client to pull my mail over dialup to mark it as spam so I could see if there was anything remotely useful.
People weren't so pissed off when they'd get one or two items of junk mail mixed in with all of the other email in their INBOXes. Nowadays, it's all inverted and we have to flip through our SPAMBOX looking for anything that might be from a human.
But the fact remains that no one has ever paid me (or anyone else I know) to write code in Ocaml, Haskel, Oberon, Prolog, or D.
I'm paid to bring a concept to life, not write code in a particular langauge. I have written code at my current company in Java, ocaml, python, C, sh, scheme, and even a bit of perl (that was maintenence, though). I had written some Eiffel, too, but it didn't get deployed because it didn't seem to run nearly as fast as the Eiffel equivalent.
I use the tool I feel will allow me to fulfill my requirements most succesfully and quickly, and then I move on. If I make a case for writing (at least a portion of) our network management system in erlang, it will happen, and it will be successful.
I'm sitting beside a NeXT running NS 3.3. I've got dev 3.3 on it which will build a binary that will run on m68k (black hardware), x86, Sparc, and HPPA. They're checkboxes in project builder, or -arch options to cc.
So, yeah.:)
blackhole has reasonable prices. It's still about $300 for OPENSTEP. I've been trying to modernize my 3.3, but I may give up in a couple of days.
When the NeXT cube came out, the crowd I was hanging with didn't take it seriously. Even when the 'fire sale' occured and they got really cheap, everybody scoffed at it. I wouldn't mind having one now, but as a historical curiosity for my collection, and not much else.
When the Sun Starfire came out, the crowd I was hanging with didn't take it seriously. Of course, I didn't know anything about it. I'm guessing the same is true here.
A 1994 NeXT is still sufficiently more advanced (speed aside) in many ways than more ``modern'' computers. For example, I can write a GUI application on the one in front of me in just a few minutes that can be run natively on a m68k (black hardware), x86, HPPA or Sparc processor.
OPENSTEP took it a step further and allowed you to deploy on NT (OPENSTEP Enterprise).
There was a pretty decent market for NeXT where people needed GUI applications written quickly. Consider the following:
Slashdot Mirror
I don't think that they really mean hardware will be free - only that it will be a loss leader.
No man, it's really free! You just have to pay a lot for it and use it their way.
I try to keep backups of data on another machine at my house or in another location entirely.
I totally agree with this. I was always a bit uncomfortable leaving my house until I got all of my automated backup systems in place. Nightly DB dumps and mail tars are encrypted in chunks and rsynced off site. Every night. Every commit to my arch repository is on two machines before I get my prompt back, in another city in about four hours, and another continent about two hours after that. One time setup, and I don't lose data from theft or damage.
Take DHCP for example - damn handy system, developed by microsoft.
Whence did this misinformation originate? I had a similar statement from some MS weenie at my last company. I showed him every bit of protocol documentation I could find at the time and asked him to show me the word ``Microsoft'' in any of it. The closest I found was a windows-specific extension somewhere.
DHCP is an extenion of bootp. They didn't do that, either (see RFC 951, 1534, 1542, 2131 etc...)
pages full of mySQL errors. *sigh* I need to find something else to do.
It still amazes me that there still aren't any really killer IMAP-implementations around yet. They all seem to suffer from different problems making them annoying to use in the long run.
:)
I know what you mean. I've tried many things. I always went back to pine. Mail.app has been Good Enough(tm) to keep me away from pine most of the time (not that I've got anything against pine).
IMAP's pretty amazing, though. I'd hate to try to implement it again.
Java isn't magic here. There's also ocaml, haskell, erlang, python, and countless other languages that prevent fatal accidents while still being very pleasant to use. OCaml, in particular, would probably yield a faster, safer CVS.
Or if nothing else move some of the mail to a backup directory so the poor little imap server doesn't have to deal with YOUR pack-rat habits!
:)
Heh, actually, I have been considering that. Mail.app does need to look at all of the folders every time it connects to see if anything's changed. It's working pretty well right now, but I am considering making a new mail server. Has anything good come out since SunOS 4.1.4?
Seriously, though, keeping my mail has come in way too handy for me to consider throwing it away.
Sorry, but I couldn't let this one slide. You've obviously got a special interpretation of "fast", because I tried migrating my Eudora mailboxes to Mail, on a 1Ghz Powerbook G4.
:)
Mail CHOKED on them.
Everyone's got a story and a counter-story. I've got over 100,000 messages in IMAP (101,269 as of last night, but it goes up and down), fully synced to Mail.app (bodies and attachments) indexed for searching, and used every day. It's split over 250 mail boxes (one for each month I've sent or received email as long as I've been keeping stuff).
It's amazingly fast. It makes my mail server seem fast (Sun IPX running SunOS 4.1.4 with a custom cyrus IMAPd that supports compressed mail stores and LDAP and some other stuff).
(Sorry for all the parentheticals.
Imagine if everyone settled on FireWire for the high-speed peripheral bus, and USB never got a chance? We wouldn't have the benifits of USB, namely bus-powered devices, lower cost, support for many devices on the same bus; and then much later, high-speed USB which can finally compete with FireWire regarding bandwidth.
This makes no sense. One of the benefits of USB is bus-powered devices? Like my iPod?
Lower cost? What makes USB lower cost than firewire (cost != price)?
Many devices on the same bus? Like my video camera being controlled by my powerbook as it spools video off onto an external disk (or two)?
High-speed USB that's theoretically similar in speed to firewire being developed while the new firewire standards were being developed is a benefit? That makes the latest USB (theoretically) a little more than half the speed of the latest firewire.
I mean, I'm all for competition and stuff, but USB never seemed to be in the same space as firewire.
Subversion is centralized.
With arch, for example, I've branched projects on my laptop (with my standard replicates to a webdav server and a plain http server), made whatever changes I want in a few commits, tracked more changes from head-of-line, and eventually got my changes brought into the original project (doing a merge from my tree).
I.e. anyone, anywhere can branch the tree for any reason and track other people's changes. Additionally, you can track changes from more than one tree at a time. For example, you can track head-of-line, some guy's branch that has feature you want, another guy's branch with another feature, and a third guy's branch with a bug fix. More importantly, it's easy enough that people are actually likely to try that.
Try -O2. I believe O3 includes some optimizations which generally make things run faster, but not always. Aggressive optimization is a chancy thing.
./fibo_c 43 ./fibo_c 43 ./fibo_c 43
./fibo_ml 43
/usr/libexec/gcc/darwin/ppc/3.3/specs /usr/local/lib/ocaml
I did. If you can find a configuration that has the C version running as fast as the java version, post it (and explain it if you can...but grabbing an old version of java doesn't count).
Here's what I got (of course, pay no attention to the real time, I was doing other stuff with the machine at the time):
dustin2wti:/tmp/so 509% gcc -O3 -o fibo_c fibo.c
0.020u 0.070s 0:01.49 6.0% 0+0k 27+11io 0pf+0w
dustin2wti:/tmp/so 510%
701408733
31.470u 0.090s 0:33.50 94.2% 0+0k 0+0io 0pf+0w
dustin2wti:/tmp/so 511% gcc -O2 -o fibo_c fibo.c
0.060u 0.060s 0:00.17 70.5% 0+0k 0+10io 0pf+0w
dustin2wti:/tmp/so 512%
701408733
31.670u 0.020s 0:32.43 97.7% 0+0k 0+1io 0pf+0w
dustin2wti:/tmp/so 513% gcc -O1 -o fibo_c fibo.c
0.040u 0.050s 0:00.17 52.9% 0+0k 0+6io 0pf+0w
dustin2wti:/tmp/so 514%
701408733
31.100u 0.040s 0:35.71 87.2% 0+0k 0+0io 0pf+0w
dustin2wti:/tmp/so 515% ocamlopt -o fibo_ml fibo.ml
0.120u 0.080s 0:00.49 40.8% 0+0k 4+8io 0pf+0w
dustin2wti:/tmp/so 516%
701408733
22.910u 0.060s 0:57.17 40.1% 0+0k 0+0io 0pf+0w
dustin2wti:/tmp/so 517% javac fibo.java
0.590u 0.200s 0:03.84 20.5% 0+0k 3+10io 0pf+0w
dustin2wti:/tmp/so 518% java fibo 43
701408733
23.950u 0.200s 0:38.74 62.3% 0+0k 0+2io 0pf+0w
dustin2wti:/tmp/so 519% gcc -v
Reading specs from
Thread model: posix
gcc version 3.3 20030304 (Apple Computer, Inc. build 1640)
0.000u 0.010s 0:00.02 50.0% 0+0k 0+0io 0pf+0w
dustin2wti:/tmp/so 520% ocamlopt -v
The Objective Caml native-code compiler, version 3.07+2
Standard library directory:
0.040u 0.020s 0:00.07 85.7% 0+0k 0+0io 0pf+0w
dustin2wti:/tmp/so 521% java -version
java version "1.4.2_03"
Java(TM) 2 Runtime Environment, Standard Edition (build 1.4.2_03-117.1)
Java HotSpot(TM) Client VM (build 1.4.2-34, mixed mode)
0.100u 0.070s 0:00.24 70.8% 0+0k 0+8io 0pf+0w
To use this recursive fibonacci-algorithm to compare functional languages and imperative languages is to cheat.
I don't *believe* it works the way you described (that'd make it many times faster, not just a few percent faster (23.07u vs. 31.79u)), but the point is that dealing with every possible detail only makes it harder on the programmer and harder on the compiler to come up with a better way of doing the work.
I agree that fibonacci is a dumb comparison, but it is a top-of-my-head case where you lose a lot in C (i.e. the ``fast'' language) vs. other languages.
The other thing to remember is that Java isn't a functional language.
Not familiar with OCaml
OCaml will compile to its own bytecode or native code (i.e. not via C).
but what JVM?
java version "1.4.2_03"
Java(TM) 2 Runtime Environment, Standard Edition (build 1.4.2_03-117.1)
Java HotSpot(TM) Client VM (build 1.4.2-34, mixed mode)
(OS X)
What Compiler was the C code compiled with (with what optimization options)?
gcc version 3.3 20030304 (Apple Computer, Inc. build 1640)
With -O3
Did you try other compilers/JVMs?
I tried it after reading someone's post on orkut suggesting that the java one was faster than the gcc one. The original poster was, I believe, using Linux. I had the same results on my Mac.
Was the JVM already resident in memory on your system?
I did multiple runs of each, so stuff was probably cached.
Because people sure aren't learning them from Java. C hackers are getting harder to find.
Good riddance. I'm getting annoyed by this constant assertion that C is the only real programming. It's just inappropriate for most of what people are doing with it.
For any chunk of C code using malloc() and free(), there's a smaller and faster OCaml equivalent that's garbage collected.
For fun, go to the Great Computer Language shootout and download the java, c, and ocaml fibonacci code (same algorithm). On my machine, the ocaml version is the fastest, followed by the java version, and then the C version. Including startup time, the java version is always faster (user and real CPU time) than the C version.
I write enough C, but OCaml isn't the first language that's produced faster code than I've been able to produce with a C compiler, nor is it the first that's made it easier to write reliable apps than unreliable apps.
What I really don't understand is the way people refer to free() and malloc() as if they're some sort of fundamental elements of programming that are required to be taken seriously. You know they're library routines, right? Should I not take someone seriously who doesn't use sbrk() directly in their code?
A friend of mine is complaints@gmail.com. He thought that'd be funny.
All electronic locks I have ever seen default to unlocked when the power goes off.
Smarthome sells deadbolts in either fail-safe or fail-secure. I would guess they probably sell more fail-safe ones.
Now they have a collaboration of some description with Micro$oft; it's hard to get an ally with more punch than them
Has anyone partnered with MS in any way and come out ahead? I've certainly watched plenty of MS deals go really, really wrong. Three at my last company. The best case, it just cost us money on development. Including things like modifying our web servers so their response headers didn't indicate they were Apache because Microsoft wouldn't link to a web site that wasn't running their own software.
We lost millions and our chance at a business in the worst cases. They treated us as a pilot on our development dime with the promise of an exclusive sales channel or something. When the business proved to be succesful, they made their own. The ones that failed, they just ditched.
Yet companies turn into this reverse Robin Hood thing. Most companies, it seems, will gladly go broke giving money to MS.
I'm sitting at the X Developers' Conference right now watching a presentation by some of Sun's guys on Project Looking Glass and I have to admit that this is some pretty cool shit.
I agree, but having read the Croquet documentation from a few years ago, I wouldn't exactly call it innovative.
I mean, the documentation includes screen shots of Alice looking at a mirror showing other Wonderland creatures.
I have to think Looking Glass was, um, ``inspired'' by the squeak/croquet work.
'd hate to see how you flip out when you get junk mail in real life! Do you call up these companies and threaten to sue?
That's a bad analogy. I don't get hundreds of offensive articles of junk mail every day for each person in my house. If I did, I'd do something about it.
When I go on vacation and try to check my email, it really shows me how much time it wastes. I have server-side spam filters and client-side spam filters. The server-side stuff gets a lot of the spam out of the way, but I was sitting in a hotel a couple of days ago waiting for my client to pull my mail over dialup to mark it as spam so I could see if there was anything remotely useful.
People weren't so pissed off when they'd get one or two items of junk mail mixed in with all of the other email in their INBOXes. Nowadays, it's all inverted and we have to flip through our SPAMBOX looking for anything that might be from a human.
sin (0) is not hard at all to enter on a TI calculator. Just hit SIN 0 ENTER. Faster than 0 ENTER
On my HP48, it's just as easy to hit either 0 ENTER or 0 SIN.
C++'s biggest problem is its C legacy.
Many people have many different opinions on what C++'s biggest problem is. Not to say that that isn't a valid one.
But the fact remains that no one has ever paid me (or anyone else I know) to write code in Ocaml, Haskel, Oberon, Prolog, or D.
I'm paid to bring a concept to life, not write code in a particular langauge. I have written code at my current company in Java, ocaml, python, C, sh, scheme, and even a bit of perl (that was maintenence, though). I had written some Eiffel, too, but it didn't get deployed because it didn't seem to run nearly as fast as the Eiffel equivalent.
I use the tool I feel will allow me to fulfill my requirements most succesfully and quickly, and then I move on. If I make a case for writing (at least a portion of) our network management system in erlang, it will happen, and it will be successful.
I'm sitting beside a NeXT running NS 3.3. I've got dev 3.3 on it which will build a binary that will run on m68k (black hardware), x86, Sparc, and HPPA. They're checkboxes in project builder, or -arch options to cc.
:)
So, yeah.
blackhole has reasonable prices. It's still about $300 for OPENSTEP. I've been trying to modernize my 3.3, but I may give up in a couple of days.
When the NeXT cube came out, the crowd I was hanging with didn't take it seriously. Even when the 'fire sale' occured and they got really cheap, everybody scoffed at it. I wouldn't mind having one now, but as a historical curiosity for my collection, and not much else.
When the Sun Starfire came out, the crowd I was hanging with didn't take it seriously. Of course, I didn't know anything about it. I'm guessing the same is true here.
A 1994 NeXT is still sufficiently more advanced (speed aside) in many ways than more ``modern'' computers. For example, I can write a GUI application on the one in front of me in just a few minutes that can be run natively on a m68k (black hardware), x86, HPPA or Sparc processor.
OPENSTEP took it a step further and allowed you to deploy on NT (OPENSTEP Enterprise).
There was a pretty decent market for NeXT where people needed GUI applications written quickly. Consider the following:
NeXT Order of Business.
In many theatres, an usher comes in with a flashlight about 20 minutes into a movie to look for seating problems or disruptive people.
I've never seen such a thing, and wouldn't go to a theater where it happened (well, more than once, anyway).
If I had reason to believe I was being watched at the movies, I wouldn't be there in the first place.