In practice, nobody used the user agent header in the early days to discriminate against browsers --
They just loaded up their pages with frames and tables and then added a "Looks best in Netscape" icon.
Note that Microsoft had create a 98%-correct emulation of Netscape 3 before anybody would use their browser. Mozilla faces a similar task now.
Re:Only one person needs to do it
on
MP3Pro Released
·
· Score: 1
It is around now in WMP7. And soundcard manufacturers must cooperate or no driver on the Windows CD for them. Faced with a SAP-enabled file, the OS will just happily disable your pro card.
Agreed that this all is too much of a tech support nightmare for anyone to reasonably contemplate. But the infrastructure is live.
Just 2 on serial ports, or so the story goes - multiply that by all the mystery components in the average 486/586 machine of the era.
I agree that the ISA hardware detection bit was genius (MS actually bought it from an independant hacker) - it could find everything I threw at it except a Madge Token Rink card. However, I think it's been dropped from more recent OSes - Win2000 only picks up ISA PnP cards, and ISA itself is of course going away.
This whole business about Opera being advanced from a standards point of view is mostly bogus load of dog excrement put forth by that browser's advocate.loser community. Try the Box Acid CSS link in Opera, if you don't thinks so - the test shows up fine, but rest of the page is hosed.
If you want to be like the iCab guys and argue speed and features, that's fine. But don't argue standards with Opera.
Now, if you are going to suggest banning a browser, why suggests MSIE 6 -- if you have any influence (not that you do) that effectively keeps your customers on IE 5.5 or below -- which is much less standards compliant.
Now, if you suggested banning completely broken user agents such as Netscape 4.x, I'd be inclined to agree.
I would argue the the real cause of DLL Hell in Windows wasn't technical at all.
Microsoft faced a problem back in the early 90s of how to perform OS upgrades on Windows 3.1. This was back when most people didn't have CD-ROM drives or online access, so they made the questionable decision of letting software vendors distribute the OS upgrades via their normal installers.
The problem was that software vendors (including MS) generally are only concerned with their own warez, not the whole system. Thus you had installers which did things like overwriting newer system DLLs with the older version that the vendor QAed against, or worse, vendor-supplied system DLLs which were lying about their version number or build date to appear newer than they actually were.
The situation got so bad late in Win 3.1's lifetime that it was virutally impossible to get more than a handful of programs from different vendors installed. The solution was to start over cleanly with Windows 95, but it didn't take long until the situation got bad (but not as bad) again.
The fact that this problem started cropping up in 1992 and wasn't really resolved until 2000 (and even then, only partially, because warez still need to run on Win95,etc) just demonstrates the underlying suckyness of Microsoft's approach towards building systems in the 1990s.
Many, many Windows users were crying for a DLL Hell solution for Windows 3.1 -- I would have much rather had to install Windows 95 "Service Pack 32" and have known good base of system DLLs (from which I could report incompatibilites to the vendor) than have to play installer roulete with every program.
Yes - that is the correct fix (packaging the DLLs into a common installer), politics be damned.
Look at "Microsoft Money" - Installing in onto a virgin Win98 box no doubt requires umpteen upgraded and new DLLs. Fortunately, Microsoft has created a package where you can get many of these DLL. The package is called "Internet Explorer 5.0SP1". That pisses off Netscape-fans and the government, but, more importantly, it works.
Why? Because if MS required people to find and upgrade 32 system DLLs, nobody would ever run their software. GnuCash should do the same and just put "Requires Gnome 1.43* (*see release notes for actual library requirements)" on the webpage and build the packages that way - simply because it's cheaper and easier for users to upgrade Gnome than it is to hunt-n-shoot DLLs.
It sounds like the system is now slanted towards people who don't want to install Gnome , but end up with 90% of Gnome anyways to run Gnucash.
The orgin of the HCL was that NT 3.x was not intended to run on "standard components" at all - it was designed to run on certified systems. I'm sorta shocked it even booted on your system.
The problem is that what you thought of as "standard" components probably weren't. Was it an authentic i8042 keyboard controller, or a clone? Was it an authentic Rockwell whatever serial controller, or a clone? (Before Windows 95, was realeased, I read that Microsoft had two full time engineers doing nothing but writing detection code and working around bugs in the thousands of 'clone' serial controllers out there. Linux has taken a similar elbow-grease approach to PC hardware compatibility.)
They loosened hardware requirements for 4.0, and as a result many more people ran the OS, at the cost of greatly increasing NT's reputation for instability.
Well some Windows developers have used it - to lie about what version of the DLL they are installing in order to get the installer to drop their back-revision into the system directory.
This sorta thing is essentially why Microsoft removed control of the system DLLs from the user-err-installer.
WordPerfect driver was what? A 80x50 text mode? (I recall these real fancy XT machines that had something like a 132x80 text mode and special monitors for running 1.2.3)
The problem is that corporate IT types are a conservative lot and have loudly informed Intel that they won't buy RAMBUS.
Intel is readying a SDRAM (or DDR?)-capable PIV chipset, but I don't think it will be ready for a while after the PIV die shrink.
So, it's not the CPU - that's fine - it's the chipset that's keeping the corporate market squarely on the Pentium III. (This conservative attitude is also one of the big reasons that AMD has not cracked the corporate market - the CPUs again are fine, it's the chipsets of questionable reliability).
Well, if you run IIS on the domain controller, you have the obvious problem of computer\IUSR_comptuer becoming domain\IUSR_domain, but that can be worked around.
In most networks, the Primary domain controlling is not a very resource-intensive application. The BDCs handle the business of communicating with the workstations, and the PDC only needs to replicate the SAM every 40-or-so minutes. So, it is the perfect box to load up services on.
IIS, on the other hand, has far too many reboot scenarios to use on a box is directly relied upon for user work. That means, just practically, no file servers or client-server servers. So, if you are firewalled, why not the PDC?
As a user application, SQL Server is well behaved as you say. Historically, the problematic stuff is the Microsoft kernel-mojo software, such IIS and the file server. And Exchange, because it has to spend lots of time fighting with it's crap JET datastores.
When PacBell came to install my DSL, it was a young kid and and older guy "riding along". Good thing too, because the kid wired everything up backassward twice before getting it right, but was all too ready to punch the clock and get out of there. I got the idea that the old guy was on early retirement and was back as a consultant.
The other interesting thing from the ZDNet article is that no numbers for NT/2000 were mentioned from the Gartner study. I wonder what the split really was among the other operating systems?
The general way this works is that the headline and the executive summary are free, the details cost $$$.
IDG's not any better -- there's been complaints over the years about how they tally the sales of Mac software for example. But at least there's not common allegations that IDG is dirty as their are with Gartner (remember when Gartner published two reports at the same time, one reading "Linux is great (0.7 probability)" and the other reading "Linux sucks (0.7 probability)"?)
I'm sure that the IDG numbers are total fluff -- they probably took total Linux downloads and ass-umed that X% were going to new server installs, and worked backwards from there. The fact that the 24% number is "shipped" is especially bogus because most servers don't ship with an OS at all.
* being able to get a dictionary lookup on an arbitrary word (good for non-English speakers)
* being able to do a google search for background information on a given word or phrase
These sorts of things could (and have been) done by writing a shell extention to extend the right-click context menu, or have even been done via javascript and the browser's button bar.
I'm all for having the browser be more intelligent about the content display (see Netscape 4's "What's Related" and Mozilla's sidebar). For example if someone types a non-anchored URL into a bbs system like Slashdot, it would be great if the browser could recognize it so that when one right-clicks on it so that it could act like a hyperlink.
But having it automatically hyperlinked would be a little scary -- it all depends on the level of intrusiveness. Without seeing MS's implementation, it's hard to say, but my guess this is an attempt to one-up Netscape's Sidebar/What's Related idea, and will probably be done in a fairly similar manner in the final form.
1) Look you catamite - The Lotus/SP6 problem was Microsoft's which is why Microsoft fixed it. (It also broke FoxPro ODBC drivers. DOH!)
2) Microsoft has a long history of analyzing competitors' products and tweaking the OS to slow/break them in such a way that it won't break their own products. Lotus, WordPerfect, Real, and Netscape have all complained of this at various points in time. Denial ain't just a river in egypt.
They bought netscape for the Netscape.com portal and its users.
Netscape.com used to be one of the most visited sites on the net. Now MSN.com is in the same position. The only thing this tells us is that there's an extrodinary number of users too lazy to switch their home pages.
Now given that, if a portal service is profitable based on eyeballs, AOL has the responsibility to keep the browser up to snuff so that lazy users get directed to the portal. Judging by the marketshare, they haven't been doing a very good job at that, nor will banner hits probably ever pay for the Mozilla development effort. (And, strangely, when they revamped Netscape.com's channels, the DHTML would only work in IE!)
Word 2000 cheats is a really smart way by loading components only when needed. Sure it starts in a few seconds (even with the Office Startup shortcut removed), but if you hit certain functions such as Find and Replace, it will hang for several seconds while that component is loaded. Depending on your POV, this behavior is either good or bad in a word processor (I find it a little annoying).
But, in a browser where the user is constantly throwing up new windows, this late loading behavior should be an absolute necessity.
Don't forget that Mozilla is still in development, and as far as I can tell, real performance optimizations didn't start coming in until the.9 milestone branch. They are nearly feature complete, so the next several months of optimzaitons should improve things. (Netscape 6.0 was properly recongized as a canard, so forget it. Most of the slow launch time was due to the fact that it loaded Java on startup! Ug.)
As for Netscape 4.x, it was unrecoverable to begin with. The only thing that really worked well was the UI bits (and then on Windows only). The fact that it's less stable than Moz 0.9 after 4 years of maintenece releases should tell you something.
Zico -- your argument would be more persuasive if there was a fundamental difference between NT4/IIS4 and W2K/IIS5. But there really isn't and you are informed enough to know that.
Now, the ActiveDirectory and other security enhancements are nice, but really don't apply to a external webserver. Another concern is that Microsoft didn't decide to aim to be a real 'enterprise' level vendor until sometime after NT5/IIS5 froze. Which means that the new regime of taking security and coding issues very seriously probably won't really rear it's head until IIS6 ships. (Unix vendors went through this same process, just about 5 years earlier than MS.)
The difference is that you can have a working desktop environment without a firm component model, but you can't have a standards compliant browser without a working cache. And the pre-0.9 Mozilla cache was totally broken -- it would be like shipping KDE 1.0 with a broken "K" menu.
This can easily be tested on WinNT 4.0 and an older system (mine was a P-133, 128MB)
1) Install IE 4 w/ ActiveDesktop
2) Upgrade to IE 5.5. Just for yucks, turn 'off' web content.
3) Notice browser windows appear within a second or two.
4) Go into Add/Remove programs and remove ActiveDesktop to get the Win95-style Explorer back. (You might need to upgrade to 5.5 again after this)
5) Notice IE windows take 10 or more seconds to appear.
Slashdot Mirror
In practice, nobody used the user agent header in the early days to discriminate against browsers --
They just loaded up their pages with frames and tables and then added a "Looks best in Netscape" icon.
Note that Microsoft had create a 98%-correct emulation of Netscape 3 before anybody would use their browser. Mozilla faces a similar task now.
It is around now in WMP7. And soundcard manufacturers must cooperate or no driver on the Windows CD for them. Faced with a SAP-enabled file, the OS will just happily disable your pro card.
Agreed that this all is too much of a tech support nightmare for anyone to reasonably contemplate. But the infrastructure is live.
Just 2 on serial ports, or so the story goes - multiply that by all the mystery components in the average 486/586 machine of the era.
I agree that the ISA hardware detection bit was genius (MS actually bought it from an independant hacker) - it could find everything I threw at it except a Madge Token Rink card. However, I think it's been dropped from more recent OSes - Win2000 only picks up ISA PnP cards, and ISA itself is of course going away.
This whole business about Opera being advanced from a standards point of view is mostly bogus load of dog excrement put forth by that browser's advocate.loser community. Try the Box Acid CSS link in Opera, if you don't thinks so - the test shows up fine, but rest of the page is hosed.
If you want to be like the iCab guys and argue speed and features, that's fine. But don't argue standards with Opera.
Now, if you are going to suggest banning a browser, why suggests MSIE 6 -- if you have any influence (not that you do) that effectively keeps your customers on IE 5.5 or below -- which is much less standards compliant.
Now, if you suggested banning completely broken user agents such as Netscape 4.x, I'd be inclined to agree.
I would argue the the real cause of DLL Hell in Windows wasn't technical at all.
Microsoft faced a problem back in the early 90s of how to perform OS upgrades on Windows 3.1. This was back when most people didn't have CD-ROM drives or online access, so they made the questionable decision of letting software vendors distribute the OS upgrades via their normal installers.
The problem was that software vendors (including MS) generally are only concerned with their own warez, not the whole system. Thus you had installers which did things like overwriting newer system DLLs with the older version that the vendor QAed against, or worse, vendor-supplied system DLLs which were lying about their version number or build date to appear newer than they actually were.
The situation got so bad late in Win 3.1's lifetime that it was virutally impossible to get more than a handful of programs from different vendors installed. The solution was to start over cleanly with Windows 95, but it didn't take long until the situation got bad (but not as bad) again.
The fact that this problem started cropping up in 1992 and wasn't really resolved until 2000 (and even then, only partially, because warez still need to run on Win95,etc) just demonstrates the underlying suckyness of Microsoft's approach towards building systems in the 1990s.
Many, many Windows users were crying for a DLL Hell solution for Windows 3.1 -- I would have much rather had to install Windows 95 "Service Pack 32" and have known good base of system DLLs (from which I could report incompatibilites to the vendor) than have to play installer roulete with every program.
Yes - that is the correct fix (packaging the DLLs into a common installer), politics be damned.
Look at "Microsoft Money" - Installing in onto a virgin Win98 box no doubt requires umpteen upgraded and new DLLs. Fortunately, Microsoft has created a package where you can get many of these DLL. The package is called "Internet Explorer 5.0SP1". That pisses off Netscape-fans and the government, but, more importantly, it works.
Why? Because if MS required people to find and upgrade 32 system DLLs, nobody would ever run their software. GnuCash should do the same and just put "Requires Gnome 1.43* (*see release notes for actual library requirements)" on the webpage and build the packages that way - simply because it's cheaper and easier for users to upgrade Gnome than it is to hunt-n-shoot DLLs.
It sounds like the system is now slanted towards people who don't want to install Gnome , but end up with 90% of Gnome anyways to run Gnucash.
I did DAE rips on 2 different NT 4.0 machines. However, they were both SCSI.
(Beyond CD-ripping, NT4 did not run well on IDE boxes. It dated from the early-90s assumption that 'workstation' == SCSI)
The orgin of the HCL was that NT 3.x was not intended to run on "standard components" at all - it was designed to run on certified systems. I'm sorta shocked it even booted on your system.
The problem is that what you thought of as "standard" components probably weren't. Was it an authentic i8042 keyboard controller, or a clone? Was it an authentic Rockwell whatever serial controller, or a clone? (Before Windows 95, was realeased, I read that Microsoft had two full time engineers doing nothing but writing detection code and working around bugs in the thousands of 'clone' serial controllers out there. Linux has taken a similar elbow-grease approach to PC hardware compatibility.)
They loosened hardware requirements for 4.0, and as a result many more people ran the OS, at the cost of greatly increasing NT's reputation for instability.
Well some Windows developers have used it - to lie about what version of the DLL they are installing in order to get the installer to drop their back-revision into the system directory.
This sorta thing is essentially why Microsoft removed control of the system DLLs from the user-err-installer.
Yup, in the early 90s, fancy video cards were called "Windows Accelerators" for a reason!
WordPerfect driver was what? A 80x50 text mode? (I recall these real fancy XT machines that had something like a 132x80 text mode and special monitors for running 1.2.3)
The problem is that corporate IT types are a conservative lot and have loudly informed Intel that they won't buy RAMBUS.
Intel is readying a SDRAM (or DDR?)-capable PIV chipset, but I don't think it will be ready for a while after the PIV die shrink.
So, it's not the CPU - that's fine - it's the chipset that's keeping the corporate market squarely on the Pentium III. (This conservative attitude is also one of the big reasons that AMD has not cracked the corporate market - the CPUs again are fine, it's the chipsets of questionable reliability).
Well, if you run IIS on the domain controller, you have the obvious problem of computer\IUSR_comptuer becoming domain\IUSR_domain, but that can be worked around.
In most networks, the Primary domain controlling is not a very resource-intensive application. The BDCs handle the business of communicating with the workstations, and the PDC only needs to replicate the SAM every 40-or-so minutes. So, it is the perfect box to load up services on.
IIS, on the other hand, has far too many reboot scenarios to use on a box is directly relied upon for user work. That means, just practically, no file servers or client-server servers. So, if you are firewalled, why not the PDC?
As a user application, SQL Server is well behaved as you say. Historically, the problematic stuff is the Microsoft kernel-mojo software, such IIS and the file server. And Exchange, because it has to spend lots of time fighting with it's crap JET datastores.
When PacBell came to install my DSL, it was a young kid and and older guy "riding along". Good thing too, because the kid wired everything up backassward twice before getting it right, but was all too ready to punch the clock and get out of there. I got the idea that the old guy was on early retirement and was back as a consultant.
The other interesting thing from the ZDNet article is that no numbers for NT/2000 were mentioned from the Gartner study. I wonder what the split really was among the other operating systems?
The general way this works is that the headline and the executive summary are free, the details cost $$$.
IDG's not any better -- there's been complaints over the years about how they tally the sales of Mac software for example. But at least there's not common allegations that IDG is dirty as their are with Gartner (remember when Gartner published two reports at the same time, one reading "Linux is great (0.7 probability)" and the other reading "Linux sucks (0.7 probability)"?)
I'm sure that the IDG numbers are total fluff -- they probably took total Linux downloads and ass-umed that X% were going to new server installs, and worked backwards from there. The fact that the 24% number is "shipped" is especially bogus because most servers don't ship with an OS at all.
* being able to get a dictionary lookup on an arbitrary word (good for non-English speakers)
* being able to do a google search for background information on a given word or phrase
These sorts of things could (and have been) done by writing a shell extention to extend the right-click context menu, or have even been done via javascript and the browser's button bar.
I'm all for having the browser be more intelligent about the content display (see Netscape 4's "What's Related" and Mozilla's sidebar). For example if someone types a non-anchored URL into a bbs system like Slashdot, it would be great if the browser could recognize it so that when one right-clicks on it so that it could act like a hyperlink.
But having it automatically hyperlinked would be a little scary -- it all depends on the level of intrusiveness. Without seeing MS's implementation, it's hard to say, but my guess this is an attempt to one-up Netscape's Sidebar/What's Related idea, and will probably be done in a fairly similar manner in the final form.
A much easier approach:
1) File + About Plug-ins
2) Note path to Acrobat plug-in DLL
3) Delete dll.
You, Mr. Netscape-sympathizer, are an apologist for the proprietary web protocols and embrace-and-extend tactics.
I think if you try to learn a smidgen more Javascript than where to copy-paste from, you'll find that Mozilla and IE have very similar DOM APIs.
1) Look you catamite - The Lotus/SP6 problem was Microsoft's which is why Microsoft fixed it. (It also broke FoxPro ODBC drivers. DOH!)
2) Microsoft has a long history of analyzing competitors' products and tweaking the OS to slow/break them in such a way that it won't break their own products. Lotus, WordPerfect, Real, and Netscape have all complained of this at various points in time. Denial ain't just a river in egypt.
They bought netscape for the Netscape.com portal and its users.
Netscape.com used to be one of the most visited sites on the net. Now MSN.com is in the same position. The only thing this tells us is that there's an extrodinary number of users too lazy to switch their home pages.
Now given that, if a portal service is profitable based on eyeballs, AOL has the responsibility to keep the browser up to snuff so that lazy users get directed to the portal. Judging by the marketshare, they haven't been doing a very good job at that, nor will banner hits probably ever pay for the Mozilla development effort. (And, strangely, when they revamped Netscape.com's channels, the DHTML would only work in IE!)
Word 2000 cheats is a really smart way by loading components only when needed. Sure it starts in a few seconds (even with the Office Startup shortcut removed), but if you hit certain functions such as Find and Replace, it will hang for several seconds while that component is loaded. Depending on your POV, this behavior is either good or bad in a word processor (I find it a little annoying).
.9 milestone branch. They are nearly feature complete, so the next several months of optimzaitons should improve things. (Netscape 6.0 was properly recongized as a canard, so forget it. Most of the slow launch time was due to the fact that it loaded Java on startup! Ug.)
But, in a browser where the user is constantly throwing up new windows, this late loading behavior should be an absolute necessity.
Don't forget that Mozilla is still in development, and as far as I can tell, real performance optimizations didn't start coming in until the
As for Netscape 4.x, it was unrecoverable to begin with. The only thing that really worked well was the UI bits (and then on Windows only). The fact that it's less stable than Moz 0.9 after 4 years of maintenece releases should tell you something.
Zico -- your argument would be more persuasive if there was a fundamental difference between NT4/IIS4 and W2K/IIS5. But there really isn't and you are informed enough to know that.
Now, the ActiveDirectory and other security enhancements are nice, but really don't apply to a external webserver. Another concern is that Microsoft didn't decide to aim to be a real 'enterprise' level vendor until sometime after NT5/IIS5 froze. Which means that the new regime of taking security and coding issues very seriously probably won't really rear it's head until IIS6 ships. (Unix vendors went through this same process, just about 5 years earlier than MS.)
The difference is that you can have a working desktop environment without a firm component model, but you can't have a standards compliant browser without a working cache. And the pre-0.9 Mozilla cache was totally broken -- it would be like shipping KDE 1.0 with a broken "K" menu.
This can easily be tested on WinNT 4.0 and an older system (mine was a P-133, 128MB)
1) Install IE 4 w/ ActiveDesktop
2) Upgrade to IE 5.5. Just for yucks, turn 'off' web content.
3) Notice browser windows appear within a second or two.
4) Go into Add/Remove programs and remove ActiveDesktop to get the Win95-style Explorer back. (You might need to upgrade to 5.5 again after this)
5) Notice IE windows take 10 or more seconds to appear.