ICC is really the KAI (Kuck and Associates) compiler. Intel bought KAI because the compiler rocked; Intel didn't make it fast. It was developed by outsiders, not by Intel. Therefore I think it reflects the efforts of some serious and hard-working compiler writers, not some secret Intel tricks. Simply reading and understanding the publicly available information on Intel hardware and compiler optimization is a big task.
Yes, exactly. It's fun to explore the machine, not some playpen built by a benevolent educator. I learned on TRS-80's - when I found the video RAM I was in heaven. I remember at that time hearing about LOGO and feeling really pissed off at the condescension of it.
300 300 translate 0 10 360 { 360 div 1 1 sethsbcolor 0 0 moveto 300 0 lineto 300 -20 lineto fill 10 rotate } for
Yes, I know BASIC is more intuitive. And I'm glad I learned on BASIC - if Postscript was my first language I might be insane. But Postscript is possibly the most beautiful language.
My computers don't have parity memory, and they don't crash. I do remember lusting after Northstars though. I agree with your other points - I guess the bottom line is that the computer world is in its infancy.
I think shell reuse works because we use those tools heavily by hand, so they are near to hand when writing scripts. Unfortunately, I don't know what lesson to draw from that.
Actually, mod_perl might be the answer. I work with it, and the issues you bring up are valid but can be addressed.
(I don't know how to describe it exactly, but sometimes data that is changed between hits doesn't actually reflect as changed. It's inconsistent, at best.)
A tricky topic. It's hard to tell what data you mean - some kinds of data should be persistent and some shouldn't. A global variable could persist across HTTP requests, and yet be different in each Apache child. That's why you wouldn't use them in mod_perl, except for very specific tasks.
You should store user/session data in some coherent way, such as Apache::Session, that ties it to a cookie. You should store global data, such as a product catalog, in a database. Aside from that, I'd recommend making the httpd processes stateless.
If you want the Apache children to share a data structure, you can use various modules built atop SysV shared memory. We use IPC::MM for this. We take on the burden of serializing/deserializing data, because all the modules that do that for you end up slow or buggy. In our case, it's just split/join.
I learned mod_perl via the Eagle book (Stein/MacEachern). It explains these traps. Mod_perl does eat RAM, but this is only a problem if you don't protect your server with a proxy. The proxy stops a slow client from hogging an Apache child. As for CPU, remember that forking new processes in CGI eats CPU as well. On our mod_perl servers, CPU usage stays pretty low. Most of the time the Apache children are either waiting for hits or waiting for the backend database.
I keep seeing this comment. Let me point out some issues. An IBM buyout is SCO's happiest ending. All the shareholders would be rewarded because the share price would go up in reaction to news of the buyout. Darl McBride would be an everlasting hero to the SCO shareholders and a legend in the business world as the man who brought IBM to the table and forced them to pay through the nose for a dead company. Second, McBride and friends must have negotiated their severance packages long ago. If IBM buys SCO they will have to honor them. And you seem to think that firing these guys "teaches them a lesson" or something. Absolutely not. These are executives, not hourly employees living paycheck to paycheck. They're playing a high-stakes game with other people's money, and fully expect to land on their asses if the court doesn't go their way. IBM would lose tons of money acquiring SCO at their inflated stock price. It's IBM's last resort.
Sorry, but your mouse/elephant idea has no relevance once both sides are big enough to afford a legal battle. SCO has Boies. There's not much IBM can do with more money.
MY desktop is also pretty according to my tastes, and not the tastes of a marketing rep. See, most of us don't configure our computers to be what other people think they should be, and the fact that you would presume we do is flat out insulting.
I don't think you understood. I'm not saying that you tweaked your desktop to appeal to other people. I'm saying that most people use a desktop that was crafted by Microsoft or Apple to appeal to the majority.
Re:[fvwm] auto-sorting menus
on
fvwm Turns Ten
·
· Score: 1
That's the trouble with software rivalry - software is so malleable. Emacs has a vi mode - so why does vi[m] even still exist? Well, as a vi user, I pretty much assume that vi does a better job of being vi than Emacs does. I don't really want to conduct a big experiment and note all the shortcomings, especially since all the shortcomings would be met with some excuse like "That's the best we can do given the way Emacs processes key-events." And Windows with SFU is theoreticaly Unix - more Unix than Linux is. But I don't see too many Unix folks deliberately choosing that configuration.
My path to fluxbox ran through Afterstep. All the really cool features that initially impressed me about afterstep got on my nerves after a while, until I had shut off almost every option. At that point I realized a minimal window manager (*box, ice, aewm, etc) does a better job of being a minimal window manager than afterstep can.
My colleague recently went through a similar transition from KDE to fluxbox. While I don't know the details, I think he's more comfortable.
I could turn it around and claim that fvwm with the right add-ons (panels, configurators, file explorers) is just like KDE, but we both know it would be a house of cards.
I sense that Red Hat is moving away from users like you and me. Yes, I also run Red Hat. It's time to move on to Gentoo or Debian.
[fvwm] auto-sorting menus
on
fvwm Turns Ten
·
· Score: 1
Thanks for a well thought-out reply. I don't think, though, that calling me a troll and zealot added much to its value. I'll just address one issue now, since I think it's fairly central to our disconnect:
My computer use varies from month to month and I find it damn convenient that the applications I find myself using the most I don't have to navigate to find. Perhaps you are too blinded by your anti-microsoft zealotism to see how handy something like that can be, but to many of us it is. I WILL criticize microsoft as they have done a half assed job with their OS and application suites, but I am not so blinded that I will not give credit where credit is due.
I don't start many GUI applications. (Browser is always running, xterms launched from a keybinding.) When I do, I just type 'xfig &' or whatever. No need for a menu. My background menu launches terminals ssh'd to various servers. It's arranged hierarchically by site. I don't want the WM resorting it based on usage, because it's already sorted correctly.
In a nutshell, you like Microsoft's idea but not their implementation. Hence, KDE - a hopefully superior implementation of the Microsoft idea. I don't like Microsft's idea, and the poor implementation was just icing on the cake.
Re:Windows offers a good interface..
on
fvwm Turns Ten
·
· Score: 1
Exactly, different strokes. I'm not complaining about the existence of the Windows imitators. Maybe they are even crucial to the survival of Linux. I just want to show a different perspective to those who dismiss normal window managers as not being Microsofty enough.
I tried blackbox and gave up because it lacked just one feature - can't remember which one. I guess that's what's nice about FVWM/Afterstep/WindowMaker - they dont' just encode one guy's personality, but have enough options to suit anyone. What really sold me on fluxbox was the tabbed framesets. I think blackbox has that too.
Re:Cringley, Linus, and Christoph Hellwig
on
Today's SCO News
·
· Score: 1
Well, that's definitely how techies see it. But don't forget that IBM teamed up with SCO to develop Monterey. It didn't make much sense to me at the time, but it does show IBM's recognition of SCO's capabilities.
Why FVWM matters
on
fvwm Turns Ten
·
· Score: 5, Interesting
A lot of folks seem to think that Windows represents the pinnacle of GUI aesthetics, and that everything else (except Apple) should try to copy it. These folks look down on fvwm as "not even as good as Windows 3.1".
I don't agree. I like the Unix desktop at its most Unixy - clean, efficient andminimal. No need to waste pixels catering for an idiot when this desktop is the interface for a computer professional. But if I wanted to waste some pixels, and I have in the past, I'd waste them on stuff that looks cool to my aesthetic, not what looks reassuring to some marketer trying to soothe the average user.
If you want to understand the "real" window managers, like fvwm, Afterstep, etc., realize three things:
They aren't trying to be "as good as Windows 3.1". They're in a totally different space. Just because they run on PC hardware now doesn't mean they partake of the PC mentality. These WM's can be configured from minimal to maximal, but at maximal they express a strong aesthetic that's quite different from consumer OS's.
Forget about "user friendliness". Real WM's are delicately balanced between aesthetics and efficiency, leaving little room for user-friendliness, which means accomodation to beginners. Let beginners use Gnome/KDE if they're unwilling to learn, or learn the real stuff if they're willing. More importantly, real user friendliness requires the WM to know things about applications, the machine, etc. I prefer my WM ignorant and agnostic - a mere conduit for my actions.
Don't judge them by how they "look". They don't look like anything - they're quite user-tunable, which is half the fun. The screenshots only give hints of the scope of customization. The feeling of running a desktop that you built is completely different from the feeling you get looking at someone else's desktop.
I don't like CDE very much, but CDE is clean and technical-looking in a way that Windows isn't. Almost everyone would happily go from CDE to KDE or Gnome, but I'd feel some loss of Unix flavor.
(I've ignored the fact that fvwm works with Gnome - you could have the fvwm coolness and the Gnome user-friendliness, I guess.)
I'm currently running fluxbox at work and AfterStep at home. I like a lot of what I see in the fvwm release - it seems the good window managers are converging and adopting the best features.
I know there will always be a small group that thinks as I do, but I'm afraid we're not communicating very well. Tons of newcomers are pouring into Linux, and most of them have only seen Microsoft Windows. Therefore they're inclined to view the desktop through a Microsoft lens, even as they criticize Microsoft.
I don't like Microsoft software. I find it disgusting from concept to execution, from GUI aesthetics to file formats. I don't want anything on my machines to look like that.
I'm glad to see SCO shown to be liars in public. Unfortunately, I don't think it has much bearing on the case. The core of SCO's allegation is that IBM violated nondisclosure agreements by copying secret enterprise scalability techniques from the Monterey project into Linux. This has nothing to do with the ancestral Unix code.
Maybe IBM can get the complaint dismissed since it contains incorrect statements. But maybe SCO can argue that the misstatements don't affect the core of their argument. My understanding is that if A sues B claiming 10 causes of action, and 9 of these are without merit, B could still be held liable for the 10th.
Why burden the web with page numbers? They're an artifact of a different technology. If you have a big document, make it available in one page. Then, if convenient, make it available broken down by natural units like chapters and sections.
ACKKKK!! Will everyone PLEASE stop talking about Lynx as if people still use it!!! And if you are using it, get a clue! I have handhelds that can run Firebird or Ie just fine!
I still use it. I'm posting this comment with lynx (and writing it in vi). Slashdot is a textual site and lynx is an efficient tool for handling it. I have no interest in the fonts or colors of a textual site - I'll pick my own damn font (misc-fixed) and color (green on black).
I probably use lynx for a greater percentage of my browsing than I did in 1995. That's because the web has become increasingly infested with putrid crap like unnecessary javascript and images. On the relatively innocent web of '95 I could afford to use Netscape; today the web is too choked with rubbish. The more clever the designers become with fonts, stylesheets, tables, etc. the more useful lynx becomes to bypass all that and just read the text.
I don't think you should assume that you have more clue than lynx users.
These kernel hackers seem to think the radio is DC-to-light, and if they got a driver working they could transmit on any frequency. 802.11 is 2.401 GHz to 2.473 Ghz (US & Canada, at least.) Alan Cox says:
I talked to one vendor about this stuff and fingers crossed we will see open drivers except for the radio module. In the longer term I suspect vendors will move to signed register sets, so you can load "US 802.11g" but you can't load "police frequency, full power"
In the US, most police frequencies are mid-UHF (400 MHz area) or 800 MHz. There's not much chance of a 2.4 GHz radio interfering with that. Radio design involves tradeoffs - other things being equal, a radio covering less bandwidth has higher performance. I really doubt that 802.11 radios are capable of much bandwidth outside their designed frequency range. I think this is yet another case of software folks applying the logic of infinite capability to the physical world where capabilities are very restricted.
I can think of reasons for restricting the interface which are less dramatic. First I thought of regulating transmit power - CDMA phones regulate their power based on instructions from the base station. This increases the overall capacity of the system by reducing unneeded transmission. But that doesn't seem to be it, because a google search shows a discussion of controlling 802.11 transmit power in NetBSD..
Then there's information security/crypto export. Maybe an open driver would enable users to use stronger cryptography in generating the spreading sequence. Or maybe it's just natural corporate paranoia.
I like your site, but I didn't see any transparent-background (shaped) pngs. Last time I checked, the major browsers rendered them with a gray background. That's a show-stopper for many sites using shaped images. As I said, I don't see any major web sites using png's yet. Google, Amazon, etc. are using gif. Even bigcharts, which is dynamically generating pictures and so must be paying a painful royalty to Unisys, is using gif.
Also, on http://www.pineight.com/gba/, when viewed with Netscape, the image covers up the first word(s) of some paragraphs. For example, I see this: t (.mb) programs into one ROM for. I'm not blaming PNG rendering for this, just pointing out that on a hobby site minor bugs are common and people don't usually report them. So if you did use some shaped PNG's, and some users saw gray corners, they probably wouldn't tell you.
I agree, sometimes. But if you exclude some users, the exclusion must be clearcut, which means keeping a list of compatible browsers and checking the User-Agent. Unless the list is actively maintained, new browsers will be excluded. If you skip the User-Agent check and your site looks and acts broken with some browsers it makes you look sloppy and unprofessional, which could have repercussions beyond the immediate loss of users.
JPEG became the web format of choice for photos sometime in the early 90's, where were you?
So? The point is that Unisys's patent enforcement did not drive users away from gif to jpeg. Your rebuttal doesn't address what the poster was saying.
IE and NS support it, and while it may not be widely used, any web page designer can be assured that 90+% of their readers can view PNG images.
"Not widely used" is a pretty good paraphrase of "fringe player." And anyone designing a public web site and willing to discard 10% of customers is also a fringe player.
Slashdot Mirror
ICC is really the KAI (Kuck and Associates) compiler. Intel bought KAI because the compiler rocked; Intel didn't make it fast. It was developed by outsiders, not by Intel. Therefore I think it reflects the efforts of some serious and hard-working compiler writers, not some secret Intel tricks. Simply reading and understanding the publicly available information on Intel hardware and compiler optimization is a big task.
Yes, exactly. It's fun to explore the machine, not some playpen built by a benevolent educator. I learned on TRS-80's - when I found the video RAM I was in heaven. I remember at that time hearing about LOGO and feeling really pissed off at the condescension of it.
The gs command prompt is sort of like a BASIC prompt. Add to the drawing:
See? Results in real time. Something better:
Yes, I know BASIC is more intuitive. And I'm glad I learned on BASIC - if Postscript was my first language I might be insane. But Postscript is possibly the most beautiful language.
My computers don't have parity memory, and they don't crash. I do remember lusting after Northstars though. I agree with your other points - I guess the bottom line is that the computer world is in its infancy.
You probably know this already, and it's water under the bridge, but the Perl module GD::Graph does a nice job of this.
I think shell reuse works because we use those tools heavily by hand, so they are near to hand when writing scripts. Unfortunately, I don't know what lesson to draw from that.
A tricky topic. It's hard to tell what data you mean - some kinds of data should be persistent and some shouldn't. A global variable could persist across HTTP requests, and yet be different in each Apache child. That's why you wouldn't use them in mod_perl, except for very specific tasks.
You should store user/session data in some coherent way, such as Apache::Session, that ties it to a cookie. You should store global data, such as a product catalog, in a database. Aside from that, I'd recommend making the httpd processes stateless.
If you want the Apache children to share a data structure, you can use various modules built atop SysV shared memory. We use IPC::MM for this. We take on the burden of serializing/deserializing data, because all the modules that do that for you end up slow or buggy. In our case, it's just split/join.
I learned mod_perl via the Eagle book (Stein/MacEachern). It explains these traps. Mod_perl does eat RAM, but this is only a problem if you don't protect your server with a proxy. The proxy stops a slow client from hogging an Apache child. As for CPU, remember that forking new processes in CGI eats CPU as well. On our mod_perl servers, CPU usage stays pretty low. Most of the time the Apache children are either waiting for hits or waiting for the backend database.
I keep seeing this comment. Let me point out some issues. An IBM buyout is SCO's happiest ending. All the shareholders would be rewarded because the share price would go up in reaction to news of the buyout. Darl McBride would be an everlasting hero to the SCO shareholders and a legend in the business world as the man who brought IBM to the table and forced them to pay through the nose for a dead company. Second, McBride and friends must have negotiated their severance packages long ago. If IBM buys SCO they will have to honor them. And you seem to think that firing these guys "teaches them a lesson" or something. Absolutely not. These are executives, not hourly employees living paycheck to paycheck. They're playing a high-stakes game with other people's money, and fully expect to land on their asses if the court doesn't go their way. IBM would lose tons of money acquiring SCO at their inflated stock price. It's IBM's last resort.
Sorry, but your mouse/elephant idea has no relevance once both sides are big enough to afford a legal battle. SCO has Boies. There's not much IBM can do with more money.
Yes, you can be a big hit with Japanese co-workers by bellowing "kinjiru" at random moments.
But that could cause head injuries. Fortunately, they aren't attaching them to the top of your head, but to your car.
I don't think you understood. I'm not saying that you tweaked your desktop to appeal to other people. I'm saying that most people use a desktop that was crafted by Microsoft or Apple to appeal to the majority.
That's the trouble with software rivalry - software is so malleable. Emacs has a vi mode - so why does vi[m] even still exist? Well, as a vi user, I pretty much assume that vi does a better job of being vi than Emacs does. I don't really want to conduct a big experiment and note all the shortcomings, especially since all the shortcomings would be met with some excuse like "That's the best we can do given the way Emacs processes key-events." And Windows with SFU is theoreticaly Unix - more Unix than Linux is. But I don't see too many Unix folks deliberately choosing that configuration.
My path to fluxbox ran through Afterstep. All the really cool features that initially impressed me about afterstep got on my nerves after a while, until I had shut off almost every option. At that point I realized a minimal window manager (*box, ice, aewm, etc) does a better job of being a minimal window manager than afterstep can.
My colleague recently went through a similar transition from KDE to fluxbox. While I don't know the details, I think he's more comfortable.
I could turn it around and claim that fvwm with the right add-ons (panels, configurators, file explorers) is just like KDE, but we both know it would be a house of cards.
I sense that Red Hat is moving away from users like you and me. Yes, I also run Red Hat. It's time to move on to Gentoo or Debian.
I don't start many GUI applications. (Browser is always running, xterms launched from a keybinding.) When I do, I just type 'xfig &' or whatever. No need for a menu. My background menu launches terminals ssh'd to various servers. It's arranged hierarchically by site. I don't want the WM resorting it based on usage, because it's already sorted correctly.
In a nutshell, you like Microsoft's idea but not their implementation. Hence, KDE - a hopefully superior implementation of the Microsoft idea. I don't like Microsft's idea, and the poor implementation was just icing on the cake.
Exactly, different strokes. I'm not complaining about the existence of the Windows imitators. Maybe they are even crucial to the survival of Linux. I just want to show a different perspective to those who dismiss normal window managers as not being Microsofty enough.
I tried blackbox and gave up because it lacked just one feature - can't remember which one. I guess that's what's nice about FVWM/Afterstep/WindowMaker - they dont' just encode one guy's personality, but have enough options to suit anyone. What really sold me on fluxbox was the tabbed framesets. I think blackbox has that too.
Well, that's definitely how techies see it. But don't forget that IBM teamed up with SCO to develop Monterey. It didn't make much sense to me at the time, but it does show IBM's recognition of SCO's capabilities.
I don't agree. I like the Unix desktop at its most Unixy - clean, efficient andminimal. No need to waste pixels catering for an idiot when this desktop is the interface for a computer professional. But if I wanted to waste some pixels, and I have in the past, I'd waste them on stuff that looks cool to my aesthetic, not what looks reassuring to some marketer trying to soothe the average user.
If you want to understand the "real" window managers, like fvwm, Afterstep, etc., realize three things:
I don't like CDE very much, but CDE is clean and technical-looking in a way that Windows isn't. Almost everyone would happily go from CDE to KDE or Gnome, but I'd feel some loss of Unix flavor.
(I've ignored the fact that fvwm works with Gnome - you could have the fvwm coolness and the Gnome user-friendliness, I guess.)
I'm currently running fluxbox at work and AfterStep at home. I like a lot of what I see in the fvwm release - it seems the good window managers are converging and adopting the best features.
I know there will always be a small group that thinks as I do, but I'm afraid we're not communicating very well. Tons of newcomers are pouring into Linux, and most of them have only seen Microsoft Windows. Therefore they're inclined to view the desktop through a Microsoft lens, even as they criticize Microsoft.
I don't like Microsoft software. I find it disgusting from concept to execution, from GUI aesthetics to file formats. I don't want anything on my machines to look like that.
I'm glad to see SCO shown to be liars in public. Unfortunately, I don't think it has much bearing on the case. The core of SCO's allegation is that IBM violated nondisclosure agreements by copying secret enterprise scalability techniques from the Monterey project into Linux. This has nothing to do with the ancestral Unix code.
Maybe IBM can get the complaint dismissed since it contains incorrect statements. But maybe SCO can argue that the misstatements don't affect the core of their argument. My understanding is that if A sues B claiming 10 causes of action, and 9 of these are without merit, B could still be held liable for the 10th.
I think the canonical reply is You won't get it.
Why burden the web with page numbers? They're an artifact of a different technology. If you have a big document, make it available in one page. Then, if convenient, make it available broken down by natural units like chapters and sections.
I still use it. I'm posting this comment with lynx (and writing it in vi). Slashdot is a textual site and lynx is an efficient tool for handling it. I have no interest in the fonts or colors of a textual site - I'll pick my own damn font (misc-fixed) and color (green on black).
I probably use lynx for a greater percentage of my browsing than I did in 1995. That's because the web has become increasingly infested with putrid crap like unnecessary javascript and images. On the relatively innocent web of '95 I could afford to use Netscape; today the web is too choked with rubbish. The more clever the designers become with fonts, stylesheets, tables, etc. the more useful lynx becomes to bypass all that and just read the text.
I don't think you should assume that you have more clue than lynx users.
In the US, most police frequencies are mid-UHF (400 MHz area) or 800 MHz. There's not much chance of a 2.4 GHz radio interfering with that. Radio design involves tradeoffs - other things being equal, a radio covering less bandwidth has higher performance. I really doubt that 802.11 radios are capable of much bandwidth outside their designed frequency range. I think this is yet another case of software folks applying the logic of infinite capability to the physical world where capabilities are very restricted.
I can think of reasons for restricting the interface which are less dramatic. First I thought of regulating transmit power - CDMA phones regulate their power based on instructions from the base station. This increases the overall capacity of the system by reducing unneeded transmission. But that doesn't seem to be it, because a google search shows a discussion of controlling 802.11 transmit power in NetBSD..
Then there's information security/crypto export. Maybe an open driver would enable users to use stronger cryptography in generating the spreading sequence. Or maybe it's just natural corporate paranoia.
I like your site, but I didn't see any transparent-background (shaped) pngs. Last time I checked, the major browsers rendered them with a gray background. That's a show-stopper for many sites using shaped images. As I said, I don't see any major web sites using png's yet. Google, Amazon, etc. are using gif. Even bigcharts, which is dynamically generating pictures and so must be paying a painful royalty to Unisys, is using gif.
Also, on http://www.pineight.com/gba/, when viewed with Netscape, the image covers up the first word(s) of some paragraphs. For example, I see this: t (.mb) programs into one ROM for. I'm not blaming PNG rendering for this, just pointing out that on a hobby site minor bugs are common and people don't usually report them. So if you did use some shaped PNG's, and some users saw gray corners, they probably wouldn't tell you.
I agree, sometimes. But if you exclude some users, the exclusion must be clearcut, which means keeping a list of compatible browsers and checking the User-Agent. Unless the list is actively maintained, new browsers will be excluded. If you skip the User-Agent check and your site looks and acts broken with some browsers it makes you look sloppy and unprofessional, which could have repercussions beyond the immediate loss of users.
"Not widely used" is a pretty good paraphrase of "fringe player." And anyone designing a public web site and willing to discard 10% of customers is also a fringe player.