Can we please stop treating Java as if it's a serious language? It's a complete and total failure on so many levels. Writing an application in C, and using a cross-platform toolkit such as GTK+ could easily avoid many of the headaches of Java, whilst offering vastly improved performance and usability. You're horribly overgeneralizing.
Java and C/C++ serve completely different needs. No one is writing system software in Java, heck even the JVM is written in C++ (iirc).
But on the flip side no one is going to write a web-based enterprise financial accounting system (or just about any business app) in C or C++. It just doesnt make sense.
For places where Java can provide a solution (ie, its possible to solve the problem in Java) then Java is probably a better solution. Because it'll be done in 1/10th the time, and for a fraction of the development costs.
AND you have limited access to the code when scaling problems start happening.
You dont need access to code to scale, you just make sure you design the app right from the start, and then just scale out horizontally (more machines load balanced) or vertically (bigger machines).
Assuming your app isnt designed horribly, and is reasonably efficient, then horizontal scaling is infinite. There are no inherent scaling problems. Both windows and linux scale reasonably linearly with more processors under load. Linux is a little leaner, same with Apache, so you see that, but we're talking incremental gains, not significant ones.
Again, just like when you need to scale on a LAMP stack, you dont go in and modify Linux, Apache, MySQL code, you just make sure you've architected your system to be able to scale reasonably linearly with hardware, and you add more hardware.
Yes, there is tuning and such you do on databases, but thats true no matter what system you are on.
I guess that most engineers will understand the MS stack. Warts and all. However, I like being able to go outside the box.
The thing is, I assume you're not independently wealthy, and need to work a job to survive. And in most businesses there is rarely if ever a good business case for needing to modify the underlying operating system. I mean what a nightmare that creates, now you have to re-introduce your customizations with every patch to the kernel. You're basically managing your own custom distro at that point.
Unless there is a _very_ compelling reason to do so (google now arguably), then modifying the underlying stack (os, db, webserver) just doesnt make good sense. Thats tinkering for the sake of tinkering and creating a maintenance nightmare.
Ummm, seems like a lot of shops are jumping on the RoR or Django bandwagons because it's more productive then.Net.
They are more productive for certain kinds of solutions. But not everything. If you're developing an enterprise financial accounting system, then RoR is probably not the right choice. Now its possible that RoR for the web gui, talking to a java or.net transaction processing back-end, that may be an excellent approach. But your comparison isnt apples to apples. Dynamic/scripting languages are a good complement to strict-typed/compiled languages, I dont think they're really a competitor, other than in very specific areas (web UI for example).
I mean.Netv3 is going to get features to try to copy ruby but I have my doubts as to if it's good enough.
.NET 3.0 is already out, has been for a little while. I've got it installed on my machine now. Thats the piece that includes WPF (Windows Presentation Foundation), WCF (Windows Communication Foundation) and the Workflow state-machine stuff. Those are incredibly compelling technologies. WPF and XAML alone, for certain types of apps, is quite incredible.
I think what you're talking about with the 'try to copy ruby' is support for dynamic languages within the.NET platform. Plus a ton of features to support LINQ, which make certain aspects of most.NET languages look more dynamic than they have before. But its not about trying to copy ruby, anymore than JRuby is trying to copy ruby.
Just see what they did to all of their VB developers as soon as they latched onto.Net. (VB.net doesn't count. You can't port directly from VB6 to VB.net and the concepts are 100% different.)
They made absolutely, positively the right choice with that. VB had been running for what, 15 years at that point, and was pretty much a backwater compared to more modern languages. And its not like VB apps will stop running, the VB runtime will be part of windows for at least another 10 years.
Should they have crippled.NET so as not to have to make VB developers learn real OO? Should
Windows\System32\Inetsrv\metabase.xml
The advertise these so well, I am sure I just missed it somewhere. What advertising? Learn your tools. RTFM. This stuff is front and center in the docs for IIS.
what makes you think that they can afford the Windows Server + IIS license? Windows Server 2003 +IIS is almost as expensive as a cheap server A couple things.
1. Windows costs money, IIS does not. There is no cost for IIS, never has been. It ships as part of the OS. You can buy the absolute cheapest version of Windows 2003 server and get IIS included.
2. You can get the Windows 2003 web server version for ~$400 from NewEgg. That includes everything you need, and requires no CALs ever.
I'm not sure what kind of servers you buy, but I dont think $400 would even buy a set of redundant power supplies, much less a whole cheap server. $2000-3000 at a minimum yes, $400 no.
You must be working on IIS4/NT4 for some reason, because this isnt how IIS5 or IIS6 works. Both _have_ an admin website, but its limited to localhost only and requires local admin login to work.
On IIS6 (the version thats been shipping for nearly 5 years) the admin site does not install by default.
Now, if you want to AUTOMATE configuration of your webserver, obviously IIS royally sucks compared to Apache. Actually, automation on IIS is largely identical to Apache. They both just use text files as their configuration storage. Now if the last time you worked with IIS was in version 5, which was ~5 years ago, you wouldnt have experienced this.
But IIS6 on Win2003 server has been XML config since day one.
Plus, you can use the COM interfaces and write your own script tools, or you can use one of the many command-line tools that ship with IIS to configure it at the command line.
It seems that Microsoft itself has come to recognize the weaknesses of GUI configuration, as one of IIS 7's most notable new features is that it is configured using plain text configuration files [iis.net]. This has been true for years. IIS6 on Win2003 server is based off an XML plain-text file too. It's been this way for almost 5 years.
How can you come on this board and bitch about something that you clearly arent up to speed on the tools? Learn the system, then you gain the right to bitch about it. But dont bitch out of ignorance.
The options are there. With MS, you are stuck with the GUI, even if you want to do something via a Text file. Or you could use the text file that underlies the GUI.
Or you could use the COM interop to admin IIS.
Or you could use the command line tools that ship with IIS.
But dont let facts get in the way of your posting or anything.
You do of course realize that IIS stores its settings in a plain-text XML file, right? And you do of course realize that the GUI administration is just one of the many ways you can administer IIS, right? You can use that, you can modify the XML file, you can use one of the COM admin interfaces to administer it programmatically.
So if you dont like the damn tabs, dont use the damn tabs.
I dont know why this sort of thing keeps coming up over and over again. Nearly every windows system gives you at least 3-4 ways to administer anything:
1. GUI 2. Config file 3. COM interface (programmatic admin) 4. Command Line tool
However, what about XP? Of the MS server platforms MS Server 2003 has negligable market penetration compared to XP. This sentence makes no sense. XP is not a server platform, its a desktop.
I havent looked at any official statistics, but the vast majority of MS servers I see in the wild are 2003.
Lets see: Windows+IIS+MSSQL+.Net. each piece of that stack requires the MS tax... Only one piece of that is for-pay, and that is Windows Server.
For large heavy use production sites, you'll probably want to pay for the standard or enterprise version of MS SQL. But for most uses, including the all important developer work, you can just use sql 2005 express for free. IIS &.Net are no additional cost.
if you do what to do something that is unorthodox, you are own your own unless you have an inside to MS and can convince engineers there to pay attention to your problem. The only time this is a real problem is if you're trying to do something that isnt possible on Windows, but thats pretty rare. Just like most people dont require modification to the JVM to create a framework on the java stack, the vast majority of people dont need to be able to modify windows, or.net to be able to go outside the norm there.
I'm not saying that its the same, its just different. There are costs on both sides. For alot of mainstream business apps, the windows stack is fast and productive for many corporate shops, as you just follow the recipe. Dont have to choose between Struts, Struts Shale, JSF, Tapestry, a templating language, or raw JSP+Servlets. This is both good and bad. If what you need can fit within the mainstream practices (which applies to a very large percentage of the people out there), then its very straightforward.
For a reasonably large company, using Microsoft software and finding a show-stopping bug means that you have to switch to another software vendor, but using FLOSS means that they can fix that bug. Actually, no. For decent sized companies, or for a serious enough bug, MS just writes up a patch and gives it to you in a day or two.
I've seen it happen once, and a few times where we contacted them and it turned out to be a known bug and I got the hotfix within a few minutes.
This is fairly common in the MS world, at least for reasonable sized orgs, or partners.
A couple inaccuracies here.
3) a move from a solid, peer-reviewed codebase where users and developers actually talk to eachother to one where marketing runs everything. This is quite inaccurate. Plenty of us have open relationships with MS engineers and product managers. The latter in particular have alot of influence on how products evolve over time. Lots of professional-level MS users talk to MS developers.
4) scrapping all existing code and building everything from scratch. This is true no matter which direction you go, unless you've made a point of using platform portable tools from the get-go. So its a moot point as you get this effect no matter in which direction you move.
On the flip side there are costs as well, mostly having to do with choice. Choice can be a double-edge sword from a business perspective. Take moving from.NET to Java for your enterprise apps, for example.
On the MS stack, there is a prescriptive recipe for everything. Best practices are all up on MSDN, with some great samples and advanced guidance from the Patterns & Practices group.
On the Java side, things are much more complicated. You have at least 5 reasonable frameworks to choose from for your web UI stack. You have several to choose from on the DB interaction, even if you do ORM. Even transaction management isnt simple, and there are choices there too. Then there's the whole, should we do Tomcat and roll everything ourselves or use JBoss/Commercial container.
This much choice has a cost. It can sometimes take years working on big apps in some of these stacks to really figure out some of the deep & dirty gotchas with them. And sometimes they just expire because a competing stack wins an overwhelming mindshare (eg, Apache OJB).
How come you took that sentence of mine, that you quoted out of context? Yeah, exactly. Thanks for listening but not hearing, reading but not comprehending. It is alright, I'm not here to convert anyone, merely to find a few people I can talk to further. Fair enough, mark it up to ongoing frustration with so-called consultants out there who do such a poor job it drives me crazy, not so much aimed directly at you.
As for the Symantec issues.... *shrugs* we dont use Symantec or McAfee. I've long, long ago given up on them, after years of finding their products to be as bad or worse than much malware. Most customers I find believe us when we tell them that its cheaper in the long run to manage the systems correctly, and not give out admin rights to typical end-users. And those that dont almost always change their mind after their first very expensive cleanup at retail hourly rates.
And frankly, I'd rather get less income per customer, but have more customers with more stable environments. We have a better reputation, our workload is more predictable, and we have happier customers, all of which leads to more business.
But, I understand that not all clients are sensible. Fortunately, I've found that those we tend not to stay involved with over the years, which is probably an okay thing.
Exactly. What most USA-based media forgets is the cost of legitimate ownership on a Microsoft stack is **way** out of reach of most businesses of your usual Western economies. Do you mean non-western economies in your last sentence there? Because legitimate ownership of an MS stack is pretty easy for US companies.
XP Pro or Vista Business costs ~$50-100 per machine. Office Pro costs ~$250 per machine (lighter versions of office are much less) For small businesses, SBS costs ~$700-1000 on top of the server costs.
So thats ~$350 per person for a tool that will last at least 3 years, if not 5.
And an extra $1000 worst case for server software that does everything your business will need while its sized to work with SBS.
And you can do things cheaper if you scrimp, buy white box machines with OEM windows and cheapy/academic versions of office.
If you're an IT shop of any sort, you can become a partner for free and purchase an Action Pack for a few hundred a year that will outfit a small IT shop with everything they need.
so if working a 10 hour day to pay the CompuShop guys to defrag and virus scan your machine is what the majority want, then that's their prerogative If you were a consulting shop (I'm assuming) then shouldnt you have been setting them up and managing them properly so you didnt have to spend ANY time doing this stuff?
In a properly managed network, you dont have problems like this. The machines are built cleanly, ran by non-admin, with patches, updates, and a/v changes pushed down automatically from a central console by professionals. Logs and errors are monitored automatically, so any problems will be detected by the techs before the users.
Or was it just an easy way to make some cash flow from ignorant clients?
So are you saying that in your situation they should have sent him an un-changeable document, and then asked him to change it and put his changes in red?
That doesnt make any sense to me. This was a problem with your contract folks, not with the technology.
Sometimes documents need to be collaborated on. Sometimes sending word-processor documents over email is an okay way to do it. Mind you, its tons easier if you have them turn change tracking on, then they dont have to do anything explicit to get the changes highlighted.
Of course, no technology will protect you from malice and trickery. At some point you have to actually proof-read the contract before signing.
These are all good points, I guess we were talking about slightly different things.
For many folks, as you say, there's no compelling reason to make the move now. I would agree with that.
The point I was trying to make is that, once the teething problems are past, which is mostly one of vendors figuring out how to write stable drivers, the changes to windows in Vista are compelling, in and of themselves, and will result in a more stable system in the long run.
Just the buit-in buffer overflow protections in the stack and the protections against kernel-patching are must-haves. These are huge improvements to the system. Having a large chunk of the drivers in userspace. This kind of stuff is so huge, and so loooong overdue.
But you're right, just like XP and 2000, it takes a while for the hardware and driver situation to stabilize. I dont see any connection with the sp1 release, as most of the onus for this is on companies other than MS, but sp1 will probably be 'long enough' for most of this to start stabilizing.
And then there's the computer manufacturers. Years back, what finally got all our dell's able to come out of standby/hibernate properly ended up being bios & firmware updates. There was one particular set of those that just really stabilized the system. I assume we'll see similar patterns here.
The first one would never happen unless the organization doesnt have an IT department.
I mean, are you aware of any business that lets their CEO have a vendor-built laptop without re-imaging it, configuring it to the company standard, installing standard apps, connecting their blackberry to it, and their sprint EVDO card to it?
The first time you did that, the million and one things that you have fixed/configured in your standard build wouldnt work for him, and you'd get the same table pounding, but the demands would be to have the IT director/veep fired.
Forgive my ignorance here, but without drivers how do you specify to staple, duplex-print, hole-punch, etc on the printer?
Or fax through it, or scan from it, etc.
If all you're talking about is bare-bones no-control printing, you can do that with many printers using windows built-in drivers for the most part, though you kind of have to guess at compatible models in many cases. Not nearly as nice as CUPS mind you, but tolerably possible.
Slashdot Mirror
Java and C/C++ serve completely different needs. No one is writing system software in Java, heck even the JVM is written in C++ (iirc).
But on the flip side no one is going to write a web-based enterprise financial accounting system (or just about any business app) in C or C++. It just doesnt make sense.
For places where Java can provide a solution (ie, its possible to solve the problem in Java) then Java is probably a better solution. Because it'll be done in 1/10th the time, and for a fraction of the development costs.
AND you have limited access to the code when scaling problems start happening.
You dont need access to code to scale, you just make sure you design the app right from the start, and then just scale out horizontally (more machines load balanced) or vertically (bigger machines).
Assuming your app isnt designed horribly, and is reasonably efficient, then horizontal scaling is infinite. There are no inherent scaling problems. Both windows and linux scale reasonably linearly with more processors under load. Linux is a little leaner, same with Apache, so you see that, but we're talking incremental gains, not significant ones.
Again, just like when you need to scale on a LAMP stack, you dont go in and modify Linux, Apache, MySQL code, you just make sure you've architected your system to be able to scale reasonably linearly with hardware, and you add more hardware.
Yes, there is tuning and such you do on databases, but thats true no matter what system you are on.
I guess that most engineers will understand the MS stack. Warts and all. However, I like being able to go outside the box.
The thing is, I assume you're not independently wealthy, and need to work a job to survive. And in most businesses there is rarely if ever a good business case for needing to modify the underlying operating system. I mean what a nightmare that creates, now you have to re-introduce your customizations with every patch to the kernel. You're basically managing your own custom distro at that point.
Unless there is a _very_ compelling reason to do so (google now arguably), then modifying the underlying stack (os, db, webserver) just doesnt make good sense. Thats tinkering for the sake of tinkering and creating a maintenance nightmare.
Ummm, seems like a lot of shops are jumping on the RoR or Django bandwagons because it's more productive then .Net.
They are more productive for certain kinds of solutions. But not everything. If you're developing an enterprise financial accounting system, then RoR is probably not the right choice. Now its possible that RoR for the web gui, talking to a java or .net transaction processing back-end, that may be an excellent approach. But your comparison isnt apples to apples. Dynamic/scripting languages are a good complement to strict-typed/compiled languages, I dont think they're really a competitor, other than in very specific areas (web UI for example).
I mean .Netv3 is going to get features to try to copy ruby but I have my doubts as to if it's good enough.
I think what you're talking about with the 'try to copy ruby' is support for dynamic languages within the
Just see what they did to all of their VB developers as soon as they latched onto .Net. (VB.net doesn't count. You can't port directly from VB6 to VB.net and the concepts are 100% different.)
They made absolutely, positively the right choice with that. VB had been running for what, 15 years at that point, and was pretty much a backwater compared to more modern languages. And its not like VB apps will stop running, the VB runtime will be part of windows for at least another 10 years.
.NET so as not to have to make VB developers learn real OO? Should
Should they have crippled
You can get one central store of alot of the info from the IIS resource kit tools, and the IIS Support Center.
There's alot alot of relevant information in the IIS6 Technical Reference, available on the IIS support center linked above.
1. Windows costs money, IIS does not. There is no cost for IIS, never has been. It ships as part of the OS. You can buy the absolute cheapest version of Windows 2003 server and get IIS included.
2. You can get the Windows 2003 web server version for ~$400 from NewEgg. That includes everything you need, and requires no CALs ever.
I'm not sure what kind of servers you buy, but I dont think $400 would even buy a set of redundant power supplies, much less a whole cheap server. $2000-3000 at a minimum yes, $400 no.
You must be working on IIS4/NT4 for some reason, because this isnt how IIS5 or IIS6 works. Both _have_ an admin website, but its limited to localhost only and requires local admin login to work.
On IIS6 (the version thats been shipping for nearly 5 years) the admin site does not install by default.
But IIS6 on Win2003 server has been XML config since day one.
Plus, you can use the COM interfaces and write your own script tools, or you can use one of the many command-line tools that ship with IIS to configure it at the command line.
So if you dont like using the GUI on IIS, why dont you use the underlying XML file that the GUI is managing?
If you have to copy config from one IIS machine to another, there are simple and straightforward tools to do this.
How can you come on this board and bitch about something that you clearly arent up to speed on the tools? Learn the system, then you gain the right to bitch about it. But dont bitch out of ignorance.
Or you could use the COM interop to admin IIS.
Or you could use the command line tools that ship with IIS.
But dont let facts get in the way of your posting or anything.
You do of course realize that IIS stores its settings in a plain-text XML file, right? And you do of course realize that the GUI administration is just one of the many ways you can administer IIS, right? You can use that, you can modify the XML file, you can use one of the COM admin interfaces to administer it programmatically.
So if you dont like the damn tabs, dont use the damn tabs.
I dont know why this sort of thing keeps coming up over and over again. Nearly every windows system gives you at least 3-4 ways to administer anything:
1. GUI
2. Config file
3. COM interface (programmatic admin)
4. Command Line tool
I havent looked at any official statistics, but the vast majority of MS servers I see in the wild are 2003.
For large heavy use production sites, you'll probably want to pay for the standard or enterprise version of MS SQL. But for most uses, including the all important developer work, you can just use sql 2005 express for free. IIS &
I'm not saying that its the same, its just different. There are costs on both sides. For alot of mainstream business apps, the windows stack is fast and productive for many corporate shops, as you just follow the recipe. Dont have to choose between Struts, Struts Shale, JSF, Tapestry, a templating language, or raw JSP+Servlets. This is both good and bad. If what you need can fit within the mainstream practices (which applies to a very large percentage of the people out there), then its very straightforward.
Paul Allen may or may not have gotten screwed over by Bill & co, but he did just fine by himself.
He's now the 19th richest man in the world and the 5th richest man in America.
He didnt do too badly out of that deal.
I've seen it happen once, and a few times where we contacted them and it turned out to be a known bug and I got the hotfix within a few minutes.
This is fairly common in the MS world, at least for reasonable sized orgs, or partners.
On the flip side there are costs as well, mostly having to do with choice. Choice can be a double-edge sword from a business perspective. Take moving from
On the MS stack, there is a prescriptive recipe for everything. Best practices are all up on MSDN, with some great samples and advanced guidance from the Patterns & Practices group.
On the Java side, things are much more complicated. You have at least 5 reasonable frameworks to choose from for your web UI stack. You have several to choose from on the DB interaction, even if you do ORM. Even transaction management isnt simple, and there are choices there too. Then there's the whole, should we do Tomcat and roll everything ourselves or use JBoss/Commercial container.
This much choice has a cost. It can sometimes take years working on big apps in some of these stacks to really figure out some of the deep & dirty gotchas with them. And sometimes they just expire because a competing stack wins an overwhelming mindshare (eg, Apache OJB).
As for the Symantec issues
And frankly, I'd rather get less income per customer, but have more customers with more stable environments. We have a better reputation, our workload is more predictable, and we have happier customers, all of which leads to more business.
But, I understand that not all clients are sensible. Fortunately, I've found that those we tend not to stay involved with over the years, which is probably an okay thing.
MS funded it, its up on sf.net
http://odf-converter.sourceforge.net/
Sun has one as well, but it doesnt work with 2007.
XP Pro or Vista Business costs ~$50-100 per machine.
Office Pro costs ~$250 per machine
(lighter versions of office are much less)
For small businesses, SBS costs ~$700-1000 on top of the server costs.
So thats ~$350 per person for a tool that will last at least 3 years, if not 5.
And an extra $1000 worst case for server software that does everything your business will need while its sized to work with SBS.
And you can do things cheaper if you scrimp, buy white box machines with OEM windows and cheapy/academic versions of office.
If you're an IT shop of any sort, you can become a partner for free and purchase an Action Pack for a few hundred a year that will outfit a small IT shop with everything they need.
In a properly managed network, you dont have problems like this. The machines are built cleanly, ran by non-admin, with patches, updates, and a/v changes pushed down automatically from a central console by professionals. Logs and errors are monitored automatically, so any problems will be detected by the techs before the users.
Or was it just an easy way to make some cash flow from ignorant clients?
Fortunately, thats where all the mainstream office suites have gone nowadays.
They're all some variation on XML zipped up, with a published spec on the web to the XML.
So are you saying that in your situation they should have sent him an un-changeable document, and then asked him to change it and put his changes in red?
That doesnt make any sense to me. This was a problem with your contract folks, not with the technology.
Sometimes documents need to be collaborated on. Sometimes sending word-processor documents over email is an okay way to do it. Mind you, its tons easier if you have them turn change tracking on, then they dont have to do anything explicit to get the changes highlighted.
Of course, no technology will protect you from malice and trickery. At some point you have to actually proof-read the contract before signing.
The same is true of the current version of MS Office. Not prior versions, though.
But current is just text/xml in a zip, with a (albeit bloated) definition available up on the web on a standards site.
These are all good points, I guess we were talking about slightly different things.
For many folks, as you say, there's no compelling reason to make the move now. I would agree with that.
The point I was trying to make is that, once the teething problems are past, which is mostly one of vendors figuring out how to write stable drivers, the changes to windows in Vista are compelling, in and of themselves, and will result in a more stable system in the long run.
Just the buit-in buffer overflow protections in the stack and the protections against kernel-patching are must-haves. These are huge improvements to the system. Having a large chunk of the drivers in userspace. This kind of stuff is so huge, and so loooong overdue.
But you're right, just like XP and 2000, it takes a while for the hardware and driver situation to stabilize. I dont see any connection with the sp1 release, as most of the onus for this is on companies other than MS, but sp1 will probably be 'long enough' for most of this to start stabilizing.
And then there's the computer manufacturers. Years back, what finally got all our dell's able to come out of standby/hibernate properly ended up being bios & firmware updates. There was one particular set of those that just really stabilized the system. I assume we'll see similar patterns here.
The first one would never happen unless the organization doesnt have an IT department.
I mean, are you aware of any business that lets their CEO have a vendor-built laptop without re-imaging it, configuring it to the company standard, installing standard apps, connecting their blackberry to it, and their sprint EVDO card to it?
The first time you did that, the million and one things that you have fixed/configured in your standard build wouldnt work for him, and you'd get the same table pounding, but the demands would be to have the IT director/veep fired.
Forgive my ignorance here, but without drivers how do you specify to staple, duplex-print, hole-punch, etc on the printer?
Or fax through it, or scan from it, etc.
If all you're talking about is bare-bones no-control printing, you can do that with many printers using windows built-in drivers for the most part, though you kind of have to guess at compatible models in many cases. Not nearly as nice as CUPS mind you, but tolerably possible.