Well, Microsoft's.NET and Passport seem to be buying into Sun's silly "network is the computer" crap.
What control would I be turning over to others? If I know who I'm talking to, what I am saying, and what the other is saying, I have much more control than installing software that puts unknown things in unknown places and hides what is going on from me.
The desktop won't vanish, just like the automobile hasn't completely eliminated the horse and buggy.
There are two fundamentally different ways to perform a conditional branch. A and B are 7FFF and 8000 respectively. If A and B are unsigned, B is 1 larger than A. If A and B are signed, then B is much smaller than A. The conditional branch is done with two instructions, a compare and branch-on-condition. The difference is whether the compare or the branch instruction is the one that knows about unsigned. Since these two instructions can be in different modules, the abstract languages supportable are different. Also only one form allows for raising an exception on overflow.
The constructs in "high-level" languages are much more complicated. Most can't do coroutines or return functions.
asm hackers... convinced that there would always be a special case where they could write better code than a compiler Asm is not better than a compiler at doing things a compiler is good at. Asm is better than a compiler at doing things that a compiler cannot do. Try doing recursive coroutines in C. The compiler is a one-size-fits-all in a lot of respects. For memory management, try cases where allocated memory can move, grow or shrink dynamically and pointers stay pointing to what they are supposed to.
So far as I know That's the problem. What you know of it is that part that is open. It would be bucking historical precedent if the open standard were what Microsoft would actually use for themselves. Unless it can be demonstrated that all of Microsoft uses only what is in the open standard, then it has to be assumed that Microsoft is hiding stuff.
Somehow I trust Sun's intentions regarding Java a lot more than I trust Microsoft's intentions regarding.NET. For example, in contrast to Microsof's Passport, the Liberty Alliance is structured so that Sun cannot grab control of it.
Who cares about the desktop? The big desktop machines are becoming more and more just overgrown dumb terminals. The network is the computer. We're not there yet, but we're gaining on it. The desktop is just something to run a browser. Which browser? Should make about as much difference as which brand of ElCheapo 4 function calculator you would buy now.
I can. I've seen them. I've worked in them.
Keypunches, sorting machines, tabulating equipment like IBM 604's. The cluster almost made what would today be a poor old computer.
Columns 73 to 80 were expected to be used for some combination of deck-id and sequence. To get the deck id just run the deck through a duplicating keypunch with the right program card and the deck id in columns 73 thru whatever.
JCL syntax is almost exactly Macro Assembler syntax. That extra space is the delimiter for the comments.
JCL is used to fill in information missing from the program's DCBs (Data Control Block). Conflicting information in the JCL is ignored.
If you want to understand what is going on, see if you can lay hands on documentation for Read Job File Control Block (RDJFCB IIRC). It should help explain peculiarities such as why SYSOUT is a DISPOSITION (like KEEP or DELETE). JCL is NOT a script language. It might look like it ought to be interpreted. It isn't.
Also the IBM 7070 and 7074 (Was there an older 707?) were decimal.
Biquinary (2 of 5)
10,000 10-digit words.
Three signs. Plus, Minus, and Alpha.
99 Index words with both start and stop addresses.
Rather nice, actually.
Core memory? Core dump?
I'm an old fart, but "line" somehow misses the crampedness implied by "card". You can make an input line longer, but you cannot make an input card longer. Unless FORTRAN has improved drastically in the last decades, it likes only very rigidly formatted input. "Card" captures that sense whereas "line" doesn't.
IBM EBCDIC Punch card holds 80 bytes of data, 80 columns by 12 rows. From top to bottom, rows are 12, 11, and 0 through 9.
Numeric 0-9 (Hex F0-F9) is just single punch in the appropriate row.
Upper Case A-I (Hex C1-C9) is 12 plus 1-9
Upper Case J-R (Hex D1-D9) is 11 plus 1-9
Upper Case S-Z (Hex E2-E9) is 0 plus 2-9
Lower Case a-i (Hex 81-89) is 12 plus 0 plus 1-9
Lower Case j-r (Hex 91-99) is 12 plus 11 plus 1-9
Lower Case s-z (Hex A2-A9) is 11 plus 0 plus 2-9
Space (Hex 40) is no punches
Hex 00 is 12-0-1-8-9
Hex FF is 12-11-0-7-8-9
Signed Numeric is 12-punch for Plus or 11-punch for Minus. Not signed is taken to be Plus.
Quibble on terminology.
... introduce a new bug.
The bugs (plural) were always there. There was just no way that anyone "should" encounter them. There was no way that all the bugs could have been tested out of the system. The combination allows latent bugs to get together, come out of hiding, and become visible. Ideally, you fix both, but fixing either is enough to make the bugs go back into hiding.
There is no such thing as a debugged program (except possibly something by Knuth;)
You can't predict everything. You can't, but you need to. Open Source is a real effective and real cheap substitute for the difference.
It's the exactly that matters.
How could you precisely define what 'invalid traffic' is? Seems like something that would keep shifting with time.
What's the difference between 3-nines and 5-nines, other than lots of money? 99.9% is the same!
"Given enough eyeballs, all bugs are shallow."
An absurd fallacy.
Nope. You are assuming that's eyeballs just looking at the source without any runtime state information. You get everything arranged just so, and look at everything just so, and the deep bug materializes out of the mists and looks like a bug. This does take many eyeballs to stumble on the chance viewpoint that shows the bug as what it is.
Debugging does not start with the source. It ends with the source. (assuming you actually want to fix the bug;)
With Open Source and a bit of effort, it is possible to get pretty close to effectively 100% bug free. The remaining latent bugs are extremely difficult to find and can even be impossible to eliminate without introducing more bugs. You encounter a bug that no one else would encounter. Even if the patch to cure the bug you encounter creates seven latent bugs that you will not encounter, this is a good patch for you (only).
Building plans are worthless to most people. Most of the time. Still better if they exist.
Two latent bugs. With the source, it's almost as good as if the bugs didn't exist. The overall effect is getting 5-nines reliability at a cost of 3-nines reliability. Also if you are facing a scissors/rock/paper scenario, any assumption you make will be wrong is some cases.
For most people, most things, most of the time, source code is useless. For most people, 5-nines reliability is useless expense.
Slashdot Mirror
Well, Microsoft's .NET and Passport seem to be buying into Sun's silly "network is the computer" crap.
What control would I be turning over to others? If I know who I'm talking to, what I am saying, and what the other is saying, I have much more control than installing software that puts unknown things in unknown places and hides what is going on from me.
The desktop won't vanish, just like the automobile hasn't completely eliminated the horse and buggy.
There are two fundamentally different ways to perform a conditional branch. A and B are 7FFF and 8000 respectively. If A and B are unsigned, B is 1 larger than A. If A and B are signed, then B is much smaller than A. The conditional branch is done with two instructions, a compare and branch-on-condition. The difference is whether the compare or the branch instruction is the one that knows about unsigned. Since these two instructions can be in different modules, the abstract languages supportable are different. Also only one form allows for raising an exception on overflow.
The constructs in "high-level" languages are much more complicated. Most can't do coroutines or return functions.
asm hackers ... convinced that there would always be a special case where they could write better code than a compiler
Asm is not better than a compiler at doing things a compiler is good at. Asm is better than a compiler at doing things that a compiler cannot do. Try doing recursive coroutines in C. The compiler is a one-size-fits-all in a lot of respects. For memory management, try cases where allocated memory can move, grow or shrink dynamically and pointers stay pointing to what they are supposed to.
So far as I know
That's the problem. What you know of it is that part that is open. It would be bucking historical precedent if the open standard were what Microsoft would actually use for themselves. Unless it can be demonstrated that all of Microsoft uses only what is in the open standard, then it has to be assumed that Microsoft is hiding stuff.
Somehow I trust Sun's intentions regarding Java a lot more than I trust Microsoft's intentions regarding .NET. For example, in contrast to Microsof's Passport, the Liberty Alliance is structured so that Sun cannot grab control of it.
Who cares about the desktop? The big desktop machines are becoming more and more just overgrown dumb terminals. The network is the computer. We're not there yet, but we're gaining on it. The desktop is just something to run a browser. Which browser? Should make about as much difference as which brand of ElCheapo 4 function calculator you would buy now.
worms.
I don't think there is anything about Java that is "neat" or "little".
Tape will mess up a card reader even faster than jelly donuts.
You're right. I'm getting old and can't see so good anymore.
Rubout is all SEVEN holes punched. Hex FF doesn't even exist in 7-bit ASCII.
I can. I've seen them. I've worked in them.
Keypunches, sorting machines, tabulating equipment like IBM 604's. The cluster almost made what would today be a poor old computer.
Columns 73 to 80 were expected to be used for some combination of deck-id and sequence. To get the deck id just run the deck through a duplicating keypunch with the right program card and the deck id in columns 73 thru whatever.
JCL syntax is almost exactly Macro Assembler syntax. That extra space is the delimiter for the comments.
JCL is used to fill in information missing from the program's DCBs (Data Control Block). Conflicting information in the JCL is ignored.
If you want to understand what is going on, see if you can lay hands on documentation for Read Job File Control Block (RDJFCB IIRC). It should help explain peculiarities such as why SYSOUT is a DISPOSITION (like KEEP or DELETE). JCL is NOT a script language. It might look like it ought to be interpreted. It isn't.
Hey, I remember when the V7 was new and MTS went from 255-max line files to 32k-max line files.
For evil, try TSO.
Take one. Twist it. Set it back down.
After a minute or so it would jump up in the air.
Old Chinese rule. No Ringee, no Writee.
It's very easy to erase a paper tape.
Just hold down rubout. All holes punched.
Ever wonder why hex FF never gets a printable character?
Also the IBM 7070 and 7074 (Was there an older 707?) were decimal.
Biquinary (2 of 5)
10,000 10-digit words.
Three signs. Plus, Minus, and Alpha.
99 Index words with both start and stop addresses.
Rather nice, actually.
96 column punch cards.
Smaller. Neater. Never did that well. Too squarish, I think.
Core memory? Core dump?
I'm an old fart, but "line" somehow misses the crampedness implied by "card". You can make an input line longer, but you cannot make an input card longer. Unless FORTRAN has improved drastically in the last decades, it likes only very rigidly formatted input. "Card" captures that sense whereas "line" doesn't.
IBM EBCDIC Punch card holds 80 bytes of data, 80 columns by 12 rows. From top to bottom, rows are 12, 11, and 0 through 9.
Numeric 0-9 (Hex F0-F9) is just single punch in the appropriate row.
Upper Case A-I (Hex C1-C9) is 12 plus 1-9
Upper Case J-R (Hex D1-D9) is 11 plus 1-9
Upper Case S-Z (Hex E2-E9) is 0 plus 2-9
Lower Case a-i (Hex 81-89) is 12 plus 0 plus 1-9
Lower Case j-r (Hex 91-99) is 12 plus 11 plus 1-9
Lower Case s-z (Hex A2-A9) is 11 plus 0 plus 2-9
Space (Hex 40) is no punches
Hex 00 is 12-0-1-8-9
Hex FF is 12-11-0-7-8-9
Signed Numeric is 12-punch for Plus or 11-punch for Minus. Not signed is taken to be Plus.
A hole punch doesn't help too much. To be interesting you need to be able to unpunch a hole.
Quibble on terminology.
... introduce a new bug.
The bugs (plural) were always there. There was just no way that anyone "should" encounter them. There was no way that all the bugs could have been tested out of the system. The combination allows latent bugs to get together, come out of hiding, and become visible. Ideally, you fix both, but fixing either is enough to make the bugs go back into hiding.
There is no such thing as a debugged program (except possibly something by Knuth;)
You can't predict everything.
You can't, but you need to. Open Source is a real effective and real cheap substitute for the difference.
It's the exactly that matters.
How could you precisely define what 'invalid traffic' is? Seems like something that would keep shifting with time.
What's the difference between 3-nines and 5-nines, other than lots of money? 99.9% is the same!
"Given enough eyeballs, all bugs are shallow."
An absurd fallacy.
Nope. You are assuming that's eyeballs just looking at the source without any runtime state information. You get everything arranged just so, and look at everything just so, and the deep bug materializes out of the mists and looks like a bug. This does take many eyeballs to stumble on the chance viewpoint that shows the bug as what it is.
Debugging does not start with the source. It ends with the source. (assuming you actually want to fix the bug;)
I'd take odds on never.
With Open Source and a bit of effort, it is possible to get pretty close to effectively 100% bug free. The remaining latent bugs are extremely difficult to find and can even be impossible to eliminate without introducing more bugs. You encounter a bug that no one else would encounter. Even if the patch to cure the bug you encounter creates seven latent bugs that you will not encounter, this is a good patch for you (only).
Building plans are worthless to most people. Most of the time. Still better if they exist.
Two latent bugs. With the source, it's almost as good as if the bugs didn't exist. The overall effect is getting 5-nines reliability at a cost of 3-nines reliability. Also if you are facing a scissors/rock/paper scenario, any assumption you make will be wrong is some cases.
For most people, most things, most of the time, source code is useless. For most people, 5-nines reliability is useless expense.