If they really wanted to test performance, they would have used bzip2 -9. I don't even understand what they are testing. Are they testing how long it takes to start an app??? That can depend on where on the disk the files are stored. What a silly test.
Everyone go do bzcat time bash -c "linux-2.6.0-test2.tar.bz2 | bzip2 -9 >>/dev/null"... That will test how fast your system is more reliably. Two processes test the schedular a little, test pipes (IPC), test disk IO latency (bzcat reading data from the file), but mostly tests the raw speed of the CPU/optimization of libbzip.
That will limit Java's heap size to 8M. It will start and run just fine, but windows will say that you are using somewhere around 30-60M. Memory reporting under windows usually is flawed when it comes to Java. Java runs on embedded systems (I put a neural net on one once). I know it doesn't use that much RAM or CPU.
It did use a lot of CPU in my case because I was using floating point. Java forces IEEE compliant floating point math. This means don't try to test the FLOPS of a Java system:) If you have something that requires floating point math, make a C library for your floating point intensive sections. The embedded system didn't have an FPU anyhow, so it probably wouldn't have amounted to any significant gains if it wasn't constrained to IEEE.
On another side topic, are there any Servlet/JSP/J2EE containers/servers that don't require seperate VM's?
My idea recently was to implement your own security archeticture and Vhosting in Jetty, and launch them all in the same VM. This will let them share stale objects and heap space without compromising security. You couldn't hook it up to Apache like Tomcat, but Jetty is faster and smaller than Tomcat. The best solution is in the works, but it keeps getting pushed off. Shared VM will do this automatically for you(except for the VHost issue obviously). Sharing the heap and JIT of core objects will help performance and memory usage by a pretty significant amount.
I hear that Orion my do all this for you, but I've not seen it myself... who has any experience with this?
It doesn't quite work that way, but you are very close. They don't really swap votes at all. The same bill will have two provisions (one for the highway, one for the education). Then when senator C says, "Hey, what am I getting out of this?" they add a provision to give 20M$ to a company owned by a friend/campaign contributer (half the money makes it back to his campaign/wallet) and person C is happy. They keep toying with things until more than half the people agree on it, the president expresses his approval. The problem is the Senate. The House and th Senate don't get enough collaboration that it will make it through. This is where lobbeying comes in. Basically, you tack on something in the House for enough companies that they'll push it through the senate for you.
This is why you see the "line item veto" try to make it through the congress sometimes by a few of the more alturistic representatives. It usually gets shot down pretty quick because it would kill this vote swapping issue. Line item veto, campaign contrabution regulations, and lobbeying goes before the house every so often, and gets struck down pretty fast. I think the campaign funding issue made it through a few years back. Usually when issues like that come up, they set them to not go into effect until their political career is over or they are dead. By the time that happens, it'll get repealed.
This is a very pessimistic viewpoint, and I'm sure it's probably not as bad as I make it out, but it's not all unicorns and carebears that Morgan would have you believe. An honest politician may not be hard to find, but you aren't going to find a majority of them to be both honest, caring, and smart. That's why you can't just sign the check to get some of them. You pick their weak point, and attack it. If they aren't smart, you use the money to convince them that they'll be doing the right thing. If they aren't honest, you can just sign the check. I'm sure there is a science to it, or else companies wouldn't go through lobbeyists, would they? In any case, you can call BS on any point where someone says that you can have as much influence on the government without money.
No matter how good your idea is, if companies don't like it, it's not going to make it through congress for two reasons: 1) representatives are not a median of ideas as Morgan would have you believe... they are usually wealthier and have much more money invested in businesses, 2) lobbeyists will give company owners a greater vote than the average person in their district because they will pay your congressman to listen to them, or pay them to vote a certain way... all depends on how easy you can get them to change their mind.
Yeah, I've never been lied to by anyone in Washington. Maybe I should just believe the guy, that performs the job that I thought was the source of the corruption, when he claims he isn't acting immorally.
Don't ask the source of the problem what the problem is... The answer will always be some distorted version of you through some diplomatic shift. "It's not so corrupt, it's just all you paranoid, ignorant/.ers giving me a bad name!"
Honestly, is this any more than you expected? "I'm not the bad guy... I love you guys, except for the bad ones among you that are paranoid and/or stupid. To think that I am corrupt is to be stupid or paranoid. I feel your pain, but most of it's caused be other people(PATRIOT), not clients like mine(DMCA)."
Whoever modded that post "Insightful" needs to be M2'd out of ever having another modpoint. Saying it the first time in the article didn't make it true, and it certainly isn't any more insightful or true the second time around. The point is, I don't think many people are buying it.
I've watched CSPAN, read reports, read bills, have relatives that have campaigned on the conservative side. I may not know everything, but I think I've been around enough to make an opinion.
Here are my thoughts:
Lobbying (as in giving money to a representative in order to get their attention as they would have you call it) would be very much illegal if someone were to "lobby" the general public before election day. In fact, there are specific laws against this. By all means of consistency with other laws, influencing legislation with money or gifts should be illegal.
We aren't alone in hating lobbying. It's been on the floor of congress about once a year or two for several years. Voter moral is in the dumpsters because they don't feel they make any difference because no matter who they vote for (that has a chance of winning according to the poles), they will just be bought off. This guy doesn't want to loose his job, and he has a head on his shoulders, I think (except when it comes to his own employment), but really he is just there to take money, distribute money, all in the hopes of changing a vote. He really should only have the right to influence 2 representatives, but he has extended himself beyond the power that any one citizen should have. This is an injustice because someone else is not getting an equal amount of representation.
I can't imagine why he would try to convince us that it isn't as simple as buying votes, when voting record lines up pretty well with campaign contributions. He sited one example of very few people in a low cash deal that are exceptions to this rule. It's not like the RIAA with billions invested into campaign contributions and just about everyone voting in a particular way. Go ahead and vote against the RIAA/MPAA and see how much harder it is to get enough money to campaign next term.
The one thing that bothered me a lot other than the above was his statement on privacy. He wants to strike a balance between corporate sharing of your information and your privacy. It's information about me, and only I should be able to decide who gets it. There is no balance. They can share if they give me good enough reason to. The burden is on them to get my permission, not on me to track down everyone they gave my information to and tell them to remove it. It is just another form of harassment. If I want your product, I'll find you. The Internet makes it really easy to advertise to anyone that is looking for your services. You just throw up a website and a robots file. Everyone uses google that has an email address, and if they need hairspray, I'm sure they can find it. If that's not good enough for you, list it on EBay. Companies are looking for a cheap way out of advertising at my expense. I've made it a point to not buy anything that I get spammed with. If their widgets are so good, I'll hear it from friends.
It's been said that if you take everyone you know, and everyone they know (repeat 6-7 times) you'll end up with everyone in the world. If it's really so good that everyone likes it, it will make it to everyone. Homestarrunner.com didn't advertise, and people were quoting it on the street even before the/. story. Make something good enough to warrant attention, and you won't have to buy addresses and spam people.
Other than those two things, I like the guy. I just think his business interests are very much different than mine, and of course his opinion is going to differ when I think his business is immoral. Everyone has some justification for everything that they do. I just don't agree that paying my representative to vote against me is ethical, and that's what he gets paid to do. Want to change a vote? Take the proper route, through us, the voters. Bypassing us is not moral or ethical, and shouldn't be legal. If it worked the way he says, convincing us would be all he needed to do to get the representatives to vote the way he wanted to. If they can convince us, then why are they going around us? If they can't convince us, then they have no right to push any further.
I think the value should be measured in labor, but bounded by usefullness value.
If I work all day at bending paper clips straight, do the paper clips have more value?
The value of the items in the game can in no way be more valuable than the game itself. If you spent 20$/mo for a year, everything you have in the game summed up can be no more valuable than 240$. 240$ worth of entertainment would have been spoiled by the hackers, not 1,000$ worth of work to get the items. The usefullness of the items is set at 20$/mo worth of entertainment by your agreement with the game vendor.
Isn't this how you measure the profitability of an idea to market anyhow? You maximize supply/demand * sell price - labor. If this is >=0, then you have a good idea. The poor economic skills of those that invest 8h/d worth of "labor" into a game can't be blamed on you. They knew full well when they started playing the game that every item in the game is just a few bits on a server. They invested their time because of entertainment value, not because it was labor. They may be upset that their entertainment has been spoiled, but I don't see people sueing Lucas for 3 days worth of labor standing in line waiting for Episode 1 + 2 hours of viewing it + 7$ ticket. They chose to stand in line and watch the movie, they didn't choose for it to be bad. They can only be entitled to 7$.
I was on it already... According to the.htaccess that is on that site, this should cover them... I didn't test the IP's, so if anyone has additions or subtractions, modify this post and continue until we have a good one, because it's going on every box I have root on:) If any calls me on it, then I'll know I've helped send a signal to those bastards. When they pay artists more, charge less, or stop fighting their customers, I'll remove the ban.
iptables -A INPUT -s 12.150.191.0/24 -j REJECT --reject-with icmp-net-prohibited iptables -A INPUT -s 63.199.57.0/24 -j REJECT --reject-with icmp-net-prohibited iptables -A INPUT -s 64.166.187.0/24 -j REJECT --reject-with icmp-net-prohibited iptables -A INPUT -s 64.241.31.0/24 -j REJECT --reject-with icmp-net-prohibited iptables -A INPUT -s 65.244.101.0/24 -j REJECT --reject-with icmp-net-prohibited iptables -A INPUT -s 66.252.128.0/24 -j REJECT --reject-with icmp-net-prohibited iptables -A INPUT -s 67.112.252.0/24 -j REJECT --reject-with icmp-net-prohibited iptables -A INPUT -s 67.125.49.0/24 -j REJECT --reject-with icmp-net-prohibited iptables -A INPUT -s 81.4.78.0/24 -j REJECT --reject-with icmp-net-prohibited iptables -A INPUT -s 146.82.174.0/24 -j REJECT --reject-with icmp-net-prohibited iptables -A INPUT -s 198.70.114.0/24 -j REJECT --reject-with icmp-net-prohibited iptables -A INPUT -s 208.192.0.0/24 -j REJECT --reject-with icmp-net-prohibited iptables -A INPUT -s 208.209.2.0/24 -j REJECT --reject-with icmp-net-prohibited iptables -A INPUT -s 208.225.90.0/24 -j REJECT --reject-with icmp-net-prohibited iptables -A INPUT -s 208.229.253.0/24 -j REJECT --reject-with icmp-net-prohibited iptables -A INPUT -s 208.49.164.0/24 -j REJECT --reject-with icmp-net-prohibited iptables -A INPUT -s 208.50.66.0/24 -j REJECT --reject-with icmp-net-prohibited iptables -A INPUT -s 212.241.48.0/24 -j REJECT --reject-with icmp-net-prohibited iptables -A INPUT -s 217.228.123.0/24 -j REJECT --reject-with icmp-net-prohibited
Anyone see anything wrong with the script off the top of their head? Cross check any IP's? Are you on the list?
You can't get a patent for something that is obvious. If more than one person was working on it at the same time, and they didn't know about each other, I would classify it as an obvious solution.
Make a list of all member corperations. If you start boycotting them, you'll cut the circulation off before it gets to the head. It would probably mean no more Playstations as well though.
mysqldump does a great job backing up databases, and it does it fast. Maybe you haven't played with the CSV exporting functionality? Also, I usually write my own exporter into my own custom, more compact XML format anyhow. That way I can use XSLT and re-import the data into a massively different schema.
By your arguement, I could say I tried SQL Server 7, and it sucks, do I have to keep trying it over and over?
Consider the cost for the 41% increase. As I said before, I could make a machine for 500$ with JBoss and PostgreSQL using Linux. If I need to get higher performance, and maintain performance/dollar, I can go to IBM and get DB2 on a big iron server, and the same applications will run on it.
You're not going to convince me that performance/price ratio is better with the MS stack in the low/middle end at all. You may convince me that MS has worked their asses off, and they can compete with the big guys in terms of performance. I make a living undercutting MS shops in both service and software development. I don't walk into a small business and tell them they need to spend 6K or more on software from MS before we can even negotiate my fees above that. I don't tell people that they should work over unreliable and/or expensive internet connections in order to use an application that they rely on for all their information.
If you want to tie your application to a single vendor, that's your choice. Scalability is important because you never know how large of a market you'll end up in. Also you have to understand that making your software prohibitively expensive to the smaller people will end up costing you more money than you think. The demand for software that costs less is more (obvious economics). If I can charge 2K$ for my software, but you have to charge 7K$, and your software is 41% faster, I think I know where people will be buying their software.
I've never had a reliability problem, en masse, out of any of my clients. I don't deny that there are some sites that have bad hardware or they put their server in the middle of a the walkway, and the disk crashes. That's what backups are for.
How often have there been worms exploiting Windows + IIS + SQL Server? How often have there been worms exploiting Linux + MySQL + Apache? I think if you tally the numbers you'll find the truth about TCO. Don't give me any BS about SQL Server or IIS being more popular, because they aren't. More people are running Apache according to NetCraft.
Not to mention, since I can copy Linux how I see fit, I can just use XFS's xfsdump / | bzip2 -9 > image.xfs.bz2, burn it to a CD using Debian's default kernel and installation script edited to bzcat image.xfs.bz | xfsrestore/target and have a system up in 30 minutes of socializing + configuration time (which I can also script) vs deploying an application on windows requires hours of updates through the internet and initial installation and activation (all requiring supervision) and finally configuration. If you can patch up an NT Server system, or even install one in under 30 minutes or without supervision and legally, I would love to see it.:)
TCP benchmarks don't even include open source database servers at this point. Posting a link to that is just plain silly for comparison.
It took me all of 5 minutes to find this link (http://www.eweek.com/article2/0,3959,293,00.asp) which shows that I wasn't completely off to say unless you are using ASP, then you shouldn't use SQL Server. I'm sure your TCP benchmarks only show ASP scores anyhow.
MySQL has had several hot-backup options for quite some time. mysqldump is the most popular, but you can have real-time replication if you need to have a backup server incase the first crashes. MySQL supports FK constraints. The Beta versions support subqueries. Stored procs still have to be in C. Now view support. Supports blobs and text with GiST indexs. (I hate GiST indexes.)
There are a lot of free RDBMS's to choose from, many of them supporting subqueries.
HSQLDB:
Stored procedures in Java
Views
Triggers
Subqueries
FK constraints
Transactions
Blobs, text, unicode
McKoi:
views
subqueries
Check and FK constraints
Transactions
blobs, text, unicode
PostgreSQL:
Stored procedures in many different languages
Views
Triggers
subqueries
check and FK constraints
Transactions
Blobs, text, unicode, geological, misc., and custom data fields
GiST index
Now with better security architecture
two-phase commit and data partitioning are now under Urgent (will be in next release most likely)
SAP:
stored procedures (don't know the language)
views
triggers
subqueries (and cursors and updatable queries)
check and FK constraints
transactions (and support for transactionless tables like MySQL)
blobs, text, unicode, ebcdic
don't know about full text indexing
a lot of security architecture
good performance even with 200G of data.
Those are the ones that most people just look over for no good reason (especially the last two considering they are really great products).
I'll start to consider TCP to not be for sale when I start to see some of these RDBMS on it. I've compared, and many independant people have compared a lot of databases, and performance per price you aren't going to find anything better than MySQL and PostgreSQL.
I have set up Oracle, and it was a pain, but I didn't have to toy with it EVER again. AFAIK, Oracle, DB2, MySql, PostgreSQL, etc. have never had worms propogate through old servers. It does require someone too keep up with SQL Server more than it requires someone to keep up with any of the other alternatives. Especially those in the main package tree of debian, and it can be done remotely, easily.
Unless you are going to go all MS, then there is no performance gains from it. The same can be said for IBM/DB2, but IBM supports portability standards like J2EE. In order to use a MS product efficiently, you have to tie yourself to it, and I know very few people that are willing to do that. Especially when I can set up a MySql server for 500$ (all hardware). If you think administering Windows and SQL Server is easier than Debian, then you're way off. I ssh in to clients about once a month, or when a bug is reported, and upgrade them, usually without any intervention from me. I bet I can take my product into more businesses than you without the overhead of SQL Server and NT Server licensing.
We are upgrading the NT4 Servers... mostly to Linux. We'll probably end up buying HP's OpenMail to replace Exchange, and write replacements for the other software that requires SQL Server. It will save them money both in terms of support and licensing. ssh is much faster than Terminal Services (we use both). Our Java applications are incredibly simple to administrate. We use webstart, and put it on the server. Whenever we need to upgrade, we overwrite the files, and all the clients download it next time they need to. Since we
I just get angry thinking about all our problems with products built on SQL server, from Exchange to proprietary. That's why I've never actually developed any software for them. I don't intrinsically hate MS. I just hate MS products that build on products I've had horrible experiences with.
Have you tried to use SQL Server with any application that requires direct connections to the SQL Server running on NT4 with Windows 2000 or greater? I think the slowness of SQL Server might be caused by the same intentional slow-downs that you get when you use NT4 as a server for a newer operating system as a client. There is no disk activity on the server during the delays, and we have run packet analyzers to get to the bottom of it. We know it has nothing to do with the servers connection to the network, because we eventually replaced the server with Linux running NT4 under VMWare. Amazingly, NT4 runs faster under Linux than it did on the raw hardware for some tasks.
It's hard to say what produces the record loss problem when it occurs about once a week during heavy load(CPU maxed out). It's just a known fact to us, that we have to live with. We run integrity checks on every SQL Server in town every Friday night, and one of the two will usually find something to fix. Usually by then, the system has been brought to it's knees by the way of performance, but somehow the integrity check fixes that too. I'm assuming the integrity check reclusters the keys by recreating the tables, and then the clients don't cry on Monday morning about how slow it is. (and they do call at 8:30 about it)
I design a lot of databases, but I will admit that only a handfull (about 3) are used widely. Cross table updates and cross table deletes would have been more handy than subqueries. I have used subqueries before, but only because they were there, and I wanted to just get it done. Everytime I see a subquery I cringe because half of the time they are used inappropriatly. 49% of the time that they are used appropriatly/correctly, a good recursive join would make the algorithm Nlog(N) instead of N^2. The 1% of the time you legitimately need them, then you can complain and choose a multitude of other stable free and fast RDBMS.
I've developed stupid B2B software during the idiotic.Com era. I've helped design the database for Stock Option tracking systems (The database grows a few gigabytes a month with all the trackers.). I've developed massive CRM "solutions". (Using the buzzword because I can't think of a better way to describe anything with as many features: Inventory, Leads, Phone calls, Scripted dialing/answering, lead progress, customer satisfaction, web site content management, email routing, document management, time management and reports that link the peices in multiple ways, and all inclusively.) Those are the only projects that have needed a database that were used in large scale production. Never, in any of these, have I ever said "I need a subquery to do this." Until MySQL supported transactions, we had minor database corruption when the cleaners were set loose on the server room. With possible exception to the Stock Option tracking system, performance has never been an issue with MySQL. The stock option system is batch updated, then only queried. No need for subqueries or transactions, just fast data access. MySQL will win every time in performance when you don't need transactions.
I don't even particularly like MySQL all that much(it has a purpose that I haven't seen anything else fulfill), and I don't know why I'm even arguing about it. My preferred RDBMS has always been PostgreSQL, though I've been looking a lot at Hypersonic lately (anyone have any performance or stability stories?). Their simply isn't a good excuse to use SQL Server. Why pay so much money for nothing? If you're going to buy a database server, why not DB2 (but you have to become an IBM shop pretty much) or Oracle (If you really like to pay)? What does SQL Server do that you can't find
Re:Why are they running Windows then?
on
Can .NET Really Scale?
·
· Score: 2, Insightful
The one that is capable of reverting to Paradox does indeed use the BDE (about the only thing that uses Paradox). I happen to know quite a bit about the BDE having worked on Builder/Delphi projects in the past. If the queries are done properly, it will work in either server or client mode, but you can always try the other. I loaded up the BDE administrator, and I managed to find some set of configuration that made it a little faster, but the queries ran both server and client just fine. The sad part is, I had better results executing on client side than on server side. Running select(*) on a table should never take very long, and most of the problem, that I can tell from diagnostics, was the server just halting network traffic. It may be a bug or problem with NT4 and SQL Server, but I'm sure not going to tell them they need to spend 20K to get these applications running. I'ld much rather develop a work-alike and drive the idiots that chose SQL Server in the first place out of the market if I can. Their product doesn't work for a group of people greater than 3 despite their claims. We've sank enough debugging time into that whole thing that we could already have developed an Alpha.
It's not the rogue master browser problem, BTW. Get out any NT4 system and any operating system built on NT since, and by default, it will be hideously slow. Pull out a Windows 98 machine and put in on the same network, and it will be instant. I'm fairly certain it's intentional incompatibility that they do all the time. Running mixed iterations of MS software will cause a lot of problems. They want you to upgrade everything all the time, because that's how they squeeze more money out of you.
I'm targetting any RDBMS with a JDBC driver, not just DB2. DB2 is just a preferred upgrade path. Even with this kind of portability, I don't take too much of a performance hit. Stored procedures wouldn't help, and neither would sub-selects. The only thing to gain from upgrading beyond MySQL for this application is good realiable data partitioning/clustering. No matter what anyone tells you, don't believe them that the current hot-backup crap in MySQL is useful for anything but failover. There are no distributed transactions the last time I checked, so I can't imagine that it would be reliable if you were using transactions on two different servers simultaneously.
If you are using Java, and you want to have a good abstracted database access, you should take a look at the open source hibernate project. Excellent software, much better than CMP. It supports all kinds of automatic relationship management too. It has it's own abstracted query language that it compiles into native queries for whatever features your DBMS can handle. It's the best way to get performance out of RDBMS's without spending months developing special queries for each DBMS, and still have DBMS portability.
Alright, it's time to bring out the big clue bat, because I've heard all this before, and I've tried it, and it's just bad.
Every program a client has that uses any version of SQL Server needs constant fixing, and is incredibly slow for 7 users. I could give you a list of programs, but it's just about every program on the market. One of them has an option to use paradox database files, and we have reverted back to that because it is both faster and more stable to share the paradox file over SMB. Every week, we have to repair an Exchange database even though it has never been improperly shut down. It generally fixes it's problems, but it's slow for the 5 people that use it. There are intentional problems between NT4 and Win2K+ using NT4 as a file server with SMB. In win2k, we side stepped the problem because MS had forgot to cause problems in the IPX part of SMB, and it worked fast when we disabled TCP/IP capabilities of SMB. In XP, it seems they caught on to it, and we just switched to Samba which was much faster, plus we get to use XFS, which has never corrupted or lost a file of mine. (NTFS has eaten the entire FS to "fix" itself in the past.)
That concludes all the problems I have with MS software, other than the fact they actually paid for the system that wasn't working.
Now, lets move on to your short sighted software development views.
SQL Server is a lot closer to Access than it is to a real RDBMS server. MySQL supports transactions, and has for a long time. If you don't need sub-selects, and a great majority of software doesn't, then you'll be just fine with it. I've had to use postgresql in the past as well, and other than vacuum, which can be cron'd, it's as fast as MySQL and it supports transactions. The only reason one would need to venture away from either of these two free options is if they needed real replication. Since I'm using JDBC, instead of some stupid MS recommended way of accessing the database, I can easily just change the configuration of the application and use DB2, the smartest database server available on the market. If you use one IBM prodect, like MS, you pretty much have to buy the whole lot (it's only just more economical with IBM, not a requirement like MS). My program, currently in JBoss and MySQL, can be deployed onto WebSphere and DB2 at any point. Too bad MS SQL isn't an option because replication has never worked.
I can sell a version of my same software to a small business down the street and they don't need to spend another dime on other software. I can sell to the little people that don't need the bloat of partitioned databases. In fact, I could sell it to work on one single machine, and it will run on windows just as well as Mac or Linux. They don't even have to buy a server. I can take my same program to Mr. Fortune500 and he can go buy everything he needs from IBM or SUN or BEA or Oracle, his choice, and I can set him up with the same program, except that it can handle 1000's of users, reliably, across a cluster.
Now, lets consider my alternative of having used.Net.
I now have a peice of software that requires $5K or more of other software in order to run. Now when I take it to a small business down the street, I have to charge 5K less, in MS tax, because they just want it to work. Now, lets say I want to take it to Mr. Fortune500. He can't choose anything other than MS..Net isn't set up in a way such that my application will automatically run on a cluster. So, I can't sell it to small nor large businesses because it's a more expensive and less scalable system. See previous slashdot stories for other tales of the woes of how poorly.Net performs.
There is absolutely no way you are going to convince me that.Net was the right path. I have a larger market for my software, and it scales. MS SQL might work for some small niche of uses, but it doesn't compare to the database independance I have with Java. I choose the
Re:Why are they running Windows then?
on
Can .NET Really Scale?
·
· Score: 4, Insightful
Actually, we supply a lot of small businesses in our area with whatever tech support they need. Kind of an outsourced IT staff. Paying us to fix things is as cheap as paying an MSCE monkey to spend 8 hours to fix a 5 minute job. We support OSS, so they save on licensing too. We even have a software team to make custom software, then release it open source.
The point is, they should be looking for the right service. You don't need dedicated staff with open source software. We get a call maybe once a month about an OSS product gone bad (usually something silly that can be fixed in 5 minutes if you know what you are doing), and we ssh in and fix it. We get calls about MS products and idiots that don't turn on things before they want to use them from 8AM till close every day. I'm pretty sure that most of our clients have spent more money on MS related tech support than OSS related tech support. I can calculate right now that the TCO for a pirated MS product would still be greater than a OS product by a significant factor. The speed at which MS products have to be fixed/patched is very much greater than a properly configured Linux system, and you're paying for that hell to boot.
If you want to shoot yourself in the foot by jumping on the.Net train before you can see where the tracks are going, then you go ahead. As for me, I plan to use as much cross-platform programming (mostly Java because the GUI is the same everywhere) and free/open source software that I possibly can, mostly because the products I use like JBoss (Free J2EE), Samba, MySQL/PostgreSQL/SAP/Firebird, etc. are more stable than.Net, Windows, MSSQL, etc.
Before those of you that say the SQL Server is actually good start flaming me, that's where a lot of headaches come from. SQL Server drops records and corrupts more than MySQL before transaction support. (There, now I'll get flames from both ends.) Also consider the price you are paying. (Per connection last time I checked.) Spend more money on the hardware and get RAID-1 on good disks and a good UPS, and you will have a faster, more reliable RDBMS.
It's certainly got to be better than the bend over and take it like a machine method that this article seems to follow. Somehow I don't think a system is going to get much work done while it's being raped by some worm, if only for bandwidth reasons. Code signing is the only way to go for high security systems. Get a mathmatically verifiable hashing algorithm that would require an exponential algorithm to crack, and require multiple keys from multiple people in charge to actually sign an executable before the hardware would actually let it pass under the PC(program counter) register. That should prevent arbitrary code from every being executed. Now if you required signed data and connections, then you can spot breaches and roll back updates from those keys while still having a usable system assuming it's something that can be rolled back (not a nuclear weapon). If you have the right key to launch a nuclear weapon, then something had to have gone wrong on the real world side of things, where you will be shot for even thinking about stealing launch codes for nuclear arms.
Proper security can work. Sure there will be bugs in the software, but you better make sure that the hardware doesn't get all screwy, and you better guard the hardware physically as much as you do elsewise. If you are controling nuclear weapons, you better put as many armed guards on the box that controls the nuclear weapons as the sum of all armed guards that you have guarding each item it controls.
Best to implement two or more cleanroom implementations of the hardware so if one goofs, the other will kick in to actually have the computer "fall on the sword". It will be good to have them debug each other anyhow. Better be implemented on a tamper resistant CPU die so there is no cutting traces to avoid the security. Also you have to worry about the secrecy of both the private and public keys. It's only a matter of time before someone make a quantum computer that will give you a private key from a public of a symmetric key encryption algorithm. Yep, it would be a lot of hard work, but it's possible to build a computer that is gauranteed to be hacker proof to a calculable degree. The conclusion though would have to be that it would be easier to steel three keys to cause armageddon than actually look for a loophole.
The answer to your question can easily be reduced to, "Do you have to support it?"
Seriously though, unless you hate the people that will be stuck with the mess, tell them it will take quite some time to get a solid version out. If they shoot themselves in the foot and tell you they need something by X, then when X comes and it's ratty and buggy and they point their fingers at you, pull out the logs/archives and point right back at them. When they want you to work extra to fix a bug, be sure to point out that they caused the problem, and now they will have to hire someone else to come in and help fix it. You shouldn't be penalized for their short-sited lack of management and planning skills. I would quit if anyone set a deadline earlier than I said was possible without hiring new help. My favorite is when you tell them you could have it by X done right, and they keep bothering you with other problems and wonder why you didn't have it done on time.
If you don't want to work 60h/week patching patches, I suggest you do it right to begin with. I work in a small little IT company, and people look to us to fix their software. A lot of the time we have to tell them their software was developed by a highschool drop-out monkey, and it would take less time to rewrite the whole thing than it would take to make their software work. Usually it's something running on Access written in VBA, that would take all of 40h to replace with a decent MVC serverside scalable extendable application using free or cheap software, assuming you know all the technologies involved.
No, it's never the parents problem is it. If you want a crippled gun, get a BB gun. If you want a gun, and you don't want your children shooting each other, teach them gun safety, and LOCK the gun up, and lock the ammo in a different place. I haven't read a story about a kid being shot that didn't involve negligent parents. Why were the children in a room alone with a gun in the first place? If your children express interest in marksmanship, take them to the range and teach them gun safety. If you are too busy to take care of your children, you shouldn't have had children to begin with. But you were probably too negligent then too. Put them up for adoption. Children don't shoot themselves in the orphanage. If you're that negligent, they'll probably not notice a difference.
I'll try to point out the obvious flaws of your arguement without insulting you for not seeing just how obvious it is.
There would be more jobs if competition existed.
Businesses could spend money on industries other than software if there was a reasonable alternative to Windows or at least competition to drive prices down. (They can't get their custom software to run on anything else because MS shops are the #1 ISVs because no one can get Dell, HP, Compaq, etc. to ship alternatives because of MS threats... which should be changing, but I haven't seen it.)
So, MS is on all the desktop workstations you can get without building them yourself. Why would anyone target a platform other than Windows? They really are locked in, and this unfair practice hurts the economy, not help it. My clients can't drop all their other software and just use mine, so it has to run on windows. No one has Linux, so it doesn't have to run there. Why go through the extra work of making software work with anything but Word, when that's the only thing anyone is going to have, and they can't change because they have invested millions of dollars in software that only runs on windows because there was no alternative to even develop for when most of it was developed.
The point is, MS is making more money than they should because they are a monopoly and they set the prices on their software, and people have to pay it, or pay the costs of having all their software ported to an alternative. This is MS's power, and this is why.Net will never fully work on another system without a legal battle. MS makes way too much money blackmailing people out of upgrades with the threats of no longer supporting their OS, thus leaving them open to disasterous bugs, rampant viruses, and crackers like they have NT4. If you have NT4 servers, you have been forced into paying thousands per machine because the product you bought was defective. In a decent sized company, that can hit millions fast.
MS doesn't deserve money because they refuse to fix holes in the software you bought. The item is defective, and has only marginally worked since it was released a decade ago, so naturally, you should have to also pay for the new product too! They are a drain on the rest of the world. Everytime you buy anything from a store, part of the overhead cost you are paying for goes to MS. How much do you pay in MS tax? If only they had to cope with the prospect of people just getting Linux and running their software on it without having to port millions of lines of VB and VC++ code, and retraining everyone, and converting all their data from MS formats to another format.
Oh well, too bad MS holding a gun to your data to get money out of you is actually good for the economy, eh?
Excuse my sarcasm, I know a lot of people see it your way, but that only makes me more angry with the situation. Effeciency will help the economy, not the blindness to corperate corruption that seems to be the staple of the Republican party lately.
Re:Kiss and say goodbye to Java language!!
on
PHP 5 Beta 1
·
· Score: 1
Both you, kris, and anthony would probably do yourself well by taking a look at Jetty or Resin.
Java doesn't leak away memory unless you are running in a Solaris machine, or someone coded a class that keeps pointers to objects that should have died long ago. If you don't handle state properly, and you keep it in memory but don't periodically check to see if it expired, of course you are going to have problems.
Some JSP engines store persistance on disk, in a database, or across a cluster in either variety as well.
When people quote Tomcat and JRun as reasons why PHP is faster than JSP, I just want to shoot them in the head. At least give Jetty (open source) a chance even if you don't give Resin ($$$), BEA ($$$), or Websphere ($$$) a chance. I wouldn't trust Macromedia (JRun) to make any server-side product really... Anyone that thinks you don't need a sine function in a language to display flashy graphics obviously doesn't have much going on in the head. Tomcat was a donated reference implementation. It's primary goal until recently seems to have been stability. Now they seem to have switched gears for performance. Jetty on the other hand is appearantly programmed by a savant. Within days of the release of NIO, he seems to have reworked the entire engine to support NIO. I really don't see how anyone could not be pleased with jetty though. I set it up on a dual processor server at the local university, and classes of 15 start and restart Jetty and they don't complain about speed at all. JBoss embeds Jetty because it's so good. Jetty doesn't support working through Apache (that's what makes Tomcat slow more than anything I would say), so you may not be happy with that, but just about any project worthy of J2EE is worthy of 100$/mo dedicated hosting.
Be sure, in any case, to try different VM's to see what works best for your project. The IBM JRE seems to be the fastest, but some argue that BEA's VM (JRockit) is actually faster for server-side stuff. Both should be free, but I'm not sure of the licensing terms.
I've seen some other FUD around going the other way too. PHP's Zend compiler is a JIT compiler as well. The speed of the two should be comperable. They both have their place, but I will probably never recommend PHP for large projects because it's so hard to seperate the logic from the presentation, where there are at least 3 tools to do this with Java.
If the future of the Web is rich clients, I'm curious if PHP+Javascript+XUL will turn out better than Java Webstart. It will be interesting to see what happens, but I'm banking on Java because of the massive numbers of backers in the industry.
I hope no one got the wrong idea, but I love PHP too, and am very excited about the new version. But that one crazy AC trying to compare Java and PHP the way he did was just silly:) I have a hard time believing he was serious though, I woulda probably modded him as funny. It's sad someone actually fell for his post. There really is no speed or stability difference, it's just easier to make small sites with PHP, and large sites (or integrate with legacy systems) with J2EE.
Because java forks processes, sometimes the OS will lie about how much memory Java is actually using. Turn on the Thread count option so you can see how many threads the process has. I can run Eclipse with -Xmx4M but not 2M. When I run it with 4M, I get 30M usage in 11 threads.. 11*4=44M which would be the maximum windows could report.
If you have 11 threads, and you are using 300M, you are getting around 90% usage of a 32M heap. This means the VM has to relocate memory in order to bring up new windows, and text files you have open will be fragmented and/or garbage collected when you switch away.
Really, you should try -Xmx256M, and I bet it runs much smoother.
The default settings only lets java claim 64M of RAM last time I checked.
Isn't it funny that everyone complains about Java's memory footprint when in fact the memory usage is usually about 10% of what they think it is because Java spawns at least 4 threads, and usually 5-6. Console applications use about 5 threads and consume about 1M of RAM, but they get reported as 5M or so RAM.
By default, the OS will report each individual thread as having access to the heap, so each thread is claiming about 3M of RAM in eclipse on my computer. So, windows adds them all together (11 threads) and tells me there is 30M of RAM in use. (It will actually start with only 3M of heap on my computer, but it's slow.)
People on Linux systems add together all the VM memory usage and say Java is using 1G of RAM on my linux machine (because Java on Linux uses more RAM for some reason.) The real way to calculate the ammount of RAM in use is to take the RSS field of "ps" of the largest process (they should all be the same, but sometimes they have some handles that are just allocated to a single thread where the heap is accessable to all.).
Actually... I was running Eclipse on a Pentium 166 with 32M of RAM once. It was horribly slow because it spent most of the time swapping memory to disk. When the hard drive light wasn't constantly on, it was usable, but that situation rarely happens when X takes up all the memory already.
You can't compare versions of software that way anyhow. Software usually gets slower every revision just because. If it wasn't that way, then Windows XP would run really fast on a Pentium 166 since it is 5 generations of software more than Win95.
I don't know what you are doing to make Eclipse slow on your machine with 512M of RAM. Maybe the Webshere version just has so many plugins that you actually should set the heap size up. Are you running the latest JRE (IBM doesn't have a 1.4 for Windows the last time I checked)? I would definately say there is something wrong though if Eclipse is slow on your machine. Check how much memory it is using. If it's close to 64m, then it has filled the default maximum heap size.
You may want to play with the 256M number since you have a lot of RAM. I hope SUN comes up with a way to automatically optimize the maximum heap size dynamically.
Either way, languages are all derivatives of something else. We aren't grunting over a peice of meat, therefore keeping a language pure is idiocy to an even greater extent.
Slashdot Mirror
-Os is stupid...
/dev/null"... That will test how fast your system is more reliably. Two processes test the schedular a little, test pipes (IPC), test disk IO latency (bzcat reading data from the file), but mostly tests the raw speed of the CPU/optimization of libbzip.
Just try that on a P4.
If they really wanted to test performance, they would have used bzip2 -9. I don't even understand what they are testing. Are they testing how long it takes to start an app??? That can depend on where on the disk the files are stored. What a silly test.
Everyone go do bzcat time bash -c "linux-2.6.0-test2.tar.bz2 | bzip2 -9 >>
Eclipse will run in 4M of ram on my machine. I have to point this out every time Java is mentioned, and it's getting old.
:) If you have something that requires floating point math, make a C library for your floating point intensive sections. The embedded system didn't have an FPU anyhow, so it probably wouldn't have amounted to any significant gains if it wasn't constrained to IEEE.
Try this:
java -Xmx8M -cp startup.jar org.eclipse.core.launcher.Main
That will limit Java's heap size to 8M. It will start and run just fine, but windows will say that you are using somewhere around 30-60M. Memory reporting under windows usually is flawed when it comes to Java. Java runs on embedded systems (I put a neural net on one once). I know it doesn't use that much RAM or CPU.
It did use a lot of CPU in my case because I was using floating point. Java forces IEEE compliant floating point math. This means don't try to test the FLOPS of a Java system
On another side topic, are there any Servlet/JSP/J2EE containers/servers that don't require seperate VM's?
My idea recently was to implement your own security archeticture and Vhosting in Jetty, and launch them all in the same VM. This will let them share stale objects and heap space without compromising security. You couldn't hook it up to Apache like Tomcat, but Jetty is faster and smaller than Tomcat. The best solution is in the works, but it keeps getting pushed off. Shared VM will do this automatically for you(except for the VHost issue obviously). Sharing the heap and JIT of core objects will help performance and memory usage by a pretty significant amount.
I hear that Orion my do all this for you, but I've not seen it myself... who has any experience with this?
It doesn't quite work that way, but you are very close. They don't really swap votes at all. The same bill will have two provisions (one for the highway, one for the education). Then when senator C says, "Hey, what am I getting out of this?" they add a provision to give 20M$ to a company owned by a friend/campaign contributer (half the money makes it back to his campaign/wallet) and person C is happy. They keep toying with things until more than half the people agree on it, the president expresses his approval. The problem is the Senate. The House and th Senate don't get enough collaboration that it will make it through. This is where lobbeying comes in. Basically, you tack on something in the House for enough companies that they'll push it through the senate for you.
This is why you see the "line item veto" try to make it through the congress sometimes by a few of the more alturistic representatives. It usually gets shot down pretty quick because it would kill this vote swapping issue. Line item veto, campaign contrabution regulations, and lobbeying goes before the house every so often, and gets struck down pretty fast. I think the campaign funding issue made it through a few years back. Usually when issues like that come up, they set them to not go into effect until their political career is over or they are dead. By the time that happens, it'll get repealed.
This is a very pessimistic viewpoint, and I'm sure it's probably not as bad as I make it out, but it's not all unicorns and carebears that Morgan would have you believe. An honest politician may not be hard to find, but you aren't going to find a majority of them to be both honest, caring, and smart. That's why you can't just sign the check to get some of them. You pick their weak point, and attack it. If they aren't smart, you use the money to convince them that they'll be doing the right thing. If they aren't honest, you can just sign the check. I'm sure there is a science to it, or else companies wouldn't go through lobbeyists, would they? In any case, you can call BS on any point where someone says that you can have as much influence on the government without money.
No matter how good your idea is, if companies don't like it, it's not going to make it through congress for two reasons: 1) representatives are not a median of ideas as Morgan would have you believe... they are usually wealthier and have much more money invested in businesses, 2) lobbeyists will give company owners a greater vote than the average person in their district because they will pay your congressman to listen to them, or pay them to vote a certain way... all depends on how easy you can get them to change their mind.
Yeah, I've never been lied to by anyone in Washington. Maybe I should just believe the guy, that performs the job that I thought was the source of the corruption, when he claims he isn't acting immorally.
/.ers giving me a bad name!"
Don't ask the source of the problem what the problem is... The answer will always be some distorted version of you through some diplomatic shift. "It's not so corrupt, it's just all you paranoid, ignorant
Honestly, is this any more than you expected? "I'm not the bad guy... I love you guys, except for the bad ones among you that are paranoid and/or stupid. To think that I am corrupt is to be stupid or paranoid. I feel your pain, but most of it's caused be other people(PATRIOT), not clients like mine(DMCA)."
Whoever modded that post "Insightful" needs to be M2'd out of ever having another modpoint. Saying it the first time in the article didn't make it true, and it certainly isn't any more insightful or true the second time around. The point is, I don't think many people are buying it.
I've watched CSPAN, read reports, read bills, have relatives that have campaigned on the conservative side. I may not know everything, but I think I've been around enough to make an opinion.
/. story. Make something good enough to warrant attention, and you won't have to buy addresses and spam people.
Here are my thoughts:
Lobbying (as in giving money to a representative in order to get their attention as they would have you call it) would be very much illegal if someone were to "lobby" the general public before election day. In fact, there are specific laws against this. By all means of consistency with other laws, influencing legislation with money or gifts should be illegal.
We aren't alone in hating lobbying. It's been on the floor of congress about once a year or two for several years. Voter moral is in the dumpsters because they don't feel they make any difference because no matter who they vote for (that has a chance of winning according to the poles), they will just be bought off. This guy doesn't want to loose his job, and he has a head on his shoulders, I think (except when it comes to his own employment), but really he is just there to take money, distribute money, all in the hopes of changing a vote. He really should only have the right to influence 2 representatives, but he has extended himself beyond the power that any one citizen should have. This is an injustice because someone else is not getting an equal amount of representation.
I can't imagine why he would try to convince us that it isn't as simple as buying votes, when voting record lines up pretty well with campaign contributions. He sited one example of very few people in a low cash deal that are exceptions to this rule. It's not like the RIAA with billions invested into campaign contributions and just about everyone voting in a particular way. Go ahead and vote against the RIAA/MPAA and see how much harder it is to get enough money to campaign next term.
The one thing that bothered me a lot other than the above was his statement on privacy. He wants to strike a balance between corporate sharing of your information and your privacy. It's information about me, and only I should be able to decide who gets it. There is no balance. They can share if they give me good enough reason to. The burden is on them to get my permission, not on me to track down everyone they gave my information to and tell them to remove it. It is just another form of harassment. If I want your product, I'll find you. The Internet makes it really easy to advertise to anyone that is looking for your services. You just throw up a website and a robots file. Everyone uses google that has an email address, and if they need hairspray, I'm sure they can find it. If that's not good enough for you, list it on EBay. Companies are looking for a cheap way out of advertising at my expense. I've made it a point to not buy anything that I get spammed with. If their widgets are so good, I'll hear it from friends.
It's been said that if you take everyone you know, and everyone they know (repeat 6-7 times) you'll end up with everyone in the world. If it's really so good that everyone likes it, it will make it to everyone. Homestarrunner.com didn't advertise, and people were quoting it on the street even before the
Other than those two things, I like the guy. I just think his business interests are very much different than mine, and of course his opinion is going to differ when I think his business is immoral. Everyone has some justification for everything that they do. I just don't agree that paying my representative to vote against me is ethical, and that's what he gets paid to do. Want to change a vote? Take the proper route, through us, the voters. Bypassing us is not moral or ethical, and shouldn't be legal. If it worked the way he says, convincing us would be all he needed to do to get the representatives to vote the way he wanted to. If they can convince us, then why are they going around us? If they can't convince us, then they have no right to push any further.
I think the value should be measured in labor, but bounded by usefullness value.
If I work all day at bending paper clips straight, do the paper clips have more value?
The value of the items in the game can in no way be more valuable than the game itself. If you spent 20$/mo for a year, everything you have in the game summed up can be no more valuable than 240$. 240$ worth of entertainment would have been spoiled by the hackers, not 1,000$ worth of work to get the items. The usefullness of the items is set at 20$/mo worth of entertainment by your agreement with the game vendor.
Isn't this how you measure the profitability of an idea to market anyhow? You maximize supply/demand * sell price - labor. If this is >=0, then you have a good idea. The poor economic skills of those that invest 8h/d worth of "labor" into a game can't be blamed on you. They knew full well when they started playing the game that every item in the game is just a few bits on a server. They invested their time because of entertainment value, not because it was labor. They may be upset that their entertainment has been spoiled, but I don't see people sueing Lucas for 3 days worth of labor standing in line waiting for Episode 1 + 2 hours of viewing it + 7$ ticket. They chose to stand in line and watch the movie, they didn't choose for it to be bad. They can only be entitled to 7$.
I was on it already... .htaccess that is on that site, this should cover them... I didn't test the IP's, so if anyone has additions or subtractions, modify this post and continue until we have a good one, because it's going on every box I have root on :) If any calls me on it, then I'll know I've helped send a signal to those bastards. When they pay artists more, charge less, or stop fighting their customers, I'll remove the ban.
According to the
iptables -A INPUT -s 12.150.191.0/24 -j REJECT --reject-with icmp-net-prohibited
iptables -A INPUT -s 63.199.57.0/24 -j REJECT --reject-with icmp-net-prohibited
iptables -A INPUT -s 64.166.187.0/24 -j REJECT --reject-with icmp-net-prohibited
iptables -A INPUT -s 64.241.31.0/24 -j REJECT --reject-with icmp-net-prohibited
iptables -A INPUT -s 65.244.101.0/24 -j REJECT --reject-with icmp-net-prohibited
iptables -A INPUT -s 66.252.128.0/24 -j REJECT --reject-with icmp-net-prohibited
iptables -A INPUT -s 67.112.252.0/24 -j REJECT --reject-with icmp-net-prohibited
iptables -A INPUT -s 67.125.49.0/24 -j REJECT --reject-with icmp-net-prohibited
iptables -A INPUT -s 81.4.78.0/24 -j REJECT --reject-with icmp-net-prohibited
iptables -A INPUT -s 146.82.174.0/24 -j REJECT --reject-with icmp-net-prohibited
iptables -A INPUT -s 198.70.114.0/24 -j REJECT --reject-with icmp-net-prohibited
iptables -A INPUT -s 208.192.0.0/24 -j REJECT --reject-with icmp-net-prohibited
iptables -A INPUT -s 208.209.2.0/24 -j REJECT --reject-with icmp-net-prohibited
iptables -A INPUT -s 208.225.90.0/24 -j REJECT --reject-with icmp-net-prohibited
iptables -A INPUT -s 208.229.253.0/24 -j REJECT --reject-with icmp-net-prohibited
iptables -A INPUT -s 208.49.164.0/24 -j REJECT --reject-with icmp-net-prohibited
iptables -A INPUT -s 208.50.66.0/24 -j REJECT --reject-with icmp-net-prohibited
iptables -A INPUT -s 212.241.48.0/24 -j REJECT --reject-with icmp-net-prohibited
iptables -A INPUT -s 217.228.123.0/24 -j REJECT --reject-with icmp-net-prohibited
Anyone see anything wrong with the script off the top of their head? Cross check any IP's? Are you on the list?
You can't get a patent for something that is obvious. If more than one person was working on it at the same time, and they didn't know about each other, I would classify it as an obvious solution.
Make a list of all member corperations. If you start boycotting them, you'll cut the circulation off before it gets to the head. It would probably mean no more Playstations as well though.
mysqldump does a great job backing up databases, and it does it fast. Maybe you haven't played with the CSV exporting functionality? Also, I usually write my own exporter into my own custom, more compact XML format anyhow. That way I can use XSLT and re-import the data into a massively different schema.
/target and have a system up in 30 minutes of socializing + configuration time (which I can also script) vs deploying an application on windows requires hours of updates through the internet and initial installation and activation (all requiring supervision) and finally configuration. If you can patch up an NT Server system, or even install one in under 30 minutes or without supervision and legally, I would love to see it. :)
By your arguement, I could say I tried SQL Server 7, and it sucks, do I have to keep trying it over and over?
Consider the cost for the 41% increase. As I said before, I could make a machine for 500$ with JBoss and PostgreSQL using Linux. If I need to get higher performance, and maintain performance/dollar, I can go to IBM and get DB2 on a big iron server, and the same applications will run on it.
You're not going to convince me that performance/price ratio is better with the MS stack in the low/middle end at all. You may convince me that MS has worked their asses off, and they can compete with the big guys in terms of performance. I make a living undercutting MS shops in both service and software development. I don't walk into a small business and tell them they need to spend 6K or more on software from MS before we can even negotiate my fees above that. I don't tell people that they should work over unreliable and/or expensive internet connections in order to use an application that they rely on for all their information.
If you want to tie your application to a single vendor, that's your choice. Scalability is important because you never know how large of a market you'll end up in. Also you have to understand that making your software prohibitively expensive to the smaller people will end up costing you more money than you think. The demand for software that costs less is more (obvious economics). If I can charge 2K$ for my software, but you have to charge 7K$, and your software is 41% faster, I think I know where people will be buying their software.
I've never had a reliability problem, en masse, out of any of my clients. I don't deny that there are some sites that have bad hardware or they put their server in the middle of a the walkway, and the disk crashes. That's what backups are for.
How often have there been worms exploiting Windows + IIS + SQL Server? How often have there been worms exploiting Linux + MySQL + Apache? I think if you tally the numbers you'll find the truth about TCO. Don't give me any BS about SQL Server or IIS being more popular, because they aren't. More people are running Apache according to NetCraft.
Not to mention, since I can copy Linux how I see fit, I can just use XFS's xfsdump / | bzip2 -9 > image.xfs.bz2, burn it to a CD using Debian's default kernel and installation script edited to bzcat image.xfs.bz | xfsrestore
TCP benchmarks don't even include open source database servers at this point. Posting a link to that is just plain silly for comparison.
It took me all of 5 minutes to find this link (http://www.eweek.com/article2/0,3959,293,00.asp) which shows that I wasn't completely off to say unless you are using ASP, then you shouldn't use SQL Server. I'm sure your TCP benchmarks only show ASP scores anyhow.
MySQL has had several hot-backup options for quite some time. mysqldump is the most popular, but you can have real-time replication if you need to have a backup server incase the first crashes. MySQL supports FK constraints. The Beta versions support subqueries. Stored procs still have to be in C. Now view support. Supports blobs and text with GiST indexs. (I hate GiST indexes.)
There are a lot of free RDBMS's to choose from, many of them supporting subqueries.
HSQLDB:
Stored procedures in Java
Views
Triggers
Subqueries
FK constraints
Transactions
Blobs, text, unicode
McKoi:
views
subqueries
Check and FK constraints
Transactions
blobs, text, unicode
PostgreSQL:
Stored procedures in many different languages
Views
Triggers
subqueries
check and FK constraints
Transactions
Blobs, text, unicode, geological, misc., and custom data fields
GiST index
Now with better security architecture
two-phase commit and data partitioning are now under Urgent (will be in next release most likely)
SAP:
stored procedures (don't know the language)
views
triggers
subqueries (and cursors and updatable queries)
check and FK constraints
transactions (and support for transactionless tables like MySQL)
blobs, text, unicode, ebcdic
don't know about full text indexing
a lot of security architecture
good performance even with 200G of data.
Those are the ones that most people just look over for no good reason (especially the last two considering they are really great products).
I'll start to consider TCP to not be for sale when I start to see some of these RDBMS on it. I've compared, and many independant people have compared a lot of databases, and performance per price you aren't going to find anything better than MySQL and PostgreSQL.
I have set up Oracle, and it was a pain, but I didn't have to toy with it EVER again. AFAIK, Oracle, DB2, MySql, PostgreSQL, etc. have never had worms propogate through old servers. It does require someone too keep up with SQL Server more than it requires someone to keep up with any of the other alternatives. Especially those in the main package tree of debian, and it can be done remotely, easily.
Unless you are going to go all MS, then there is no performance gains from it. The same can be said for IBM/DB2, but IBM supports portability standards like J2EE. In order to use a MS product efficiently, you have to tie yourself to it, and I know very few people that are willing to do that. Especially when I can set up a MySql server for 500$ (all hardware). If you think administering Windows and SQL Server is easier than Debian, then you're way off. I ssh in to clients about once a month, or when a bug is reported, and upgrade them, usually without any intervention from me. I bet I can take my product into more businesses than you without the overhead of SQL Server and NT Server licensing.
We are upgrading the NT4 Servers... mostly to Linux. We'll probably end up buying HP's OpenMail to replace Exchange, and write replacements for the other software that requires SQL Server. It will save them money both in terms of support and licensing. ssh is much faster than Terminal Services (we use both). Our Java applications are incredibly simple to administrate. We use webstart, and put it on the server. Whenever we need to upgrade, we overwrite the files, and all the clients download it next time they need to. Since we
I just get angry thinking about all our problems with products built on SQL server, from Exchange to proprietary. That's why I've never actually developed any software for them. I don't intrinsically hate MS. I just hate MS products that build on products I've had horrible experiences with.
.Com era. I've helped design the database for Stock Option tracking systems (The database grows a few gigabytes a month with all the trackers.). I've developed massive CRM "solutions". (Using the buzzword because I can't think of a better way to describe anything with as many features: Inventory, Leads, Phone calls, Scripted dialing/answering, lead progress, customer satisfaction, web site content management, email routing, document management, time management and reports that link the peices in multiple ways, and all inclusively.) Those are the only projects that have needed a database that were used in large scale production. Never, in any of these, have I ever said "I need a subquery to do this." Until MySQL supported transactions, we had minor database corruption when the cleaners were set loose on the server room. With possible exception to the Stock Option tracking system, performance has never been an issue with MySQL. The stock option system is batch updated, then only queried. No need for subqueries or transactions, just fast data access. MySQL will win every time in performance when you don't need transactions.
Have you tried to use SQL Server with any application that requires direct connections to the SQL Server running on NT4 with Windows 2000 or greater? I think the slowness of SQL Server might be caused by the same intentional slow-downs that you get when you use NT4 as a server for a newer operating system as a client. There is no disk activity on the server during the delays, and we have run packet analyzers to get to the bottom of it. We know it has nothing to do with the servers connection to the network, because we eventually replaced the server with Linux running NT4 under VMWare. Amazingly, NT4 runs faster under Linux than it did on the raw hardware for some tasks.
It's hard to say what produces the record loss problem when it occurs about once a week during heavy load(CPU maxed out). It's just a known fact to us, that we have to live with. We run integrity checks on every SQL Server in town every Friday night, and one of the two will usually find something to fix. Usually by then, the system has been brought to it's knees by the way of performance, but somehow the integrity check fixes that too. I'm assuming the integrity check reclusters the keys by recreating the tables, and then the clients don't cry on Monday morning about how slow it is. (and they do call at 8:30 about it)
I design a lot of databases, but I will admit that only a handfull (about 3) are used widely. Cross table updates and cross table deletes would have been more handy than subqueries. I have used subqueries before, but only because they were there, and I wanted to just get it done. Everytime I see a subquery I cringe because half of the time they are used inappropriatly. 49% of the time that they are used appropriatly/correctly, a good recursive join would make the algorithm Nlog(N) instead of N^2. The 1% of the time you legitimately need them, then you can complain and choose a multitude of other stable free and fast RDBMS.
I've developed stupid B2B software during the idiotic
I don't even particularly like MySQL all that much(it has a purpose that I haven't seen anything else fulfill), and I don't know why I'm even arguing about it. My preferred RDBMS has always been PostgreSQL, though I've been looking a lot at Hypersonic lately (anyone have any performance or stability stories?). Their simply isn't a good excuse to use SQL Server. Why pay so much money for nothing? If you're going to buy a database server, why not DB2 (but you have to become an IBM shop pretty much) or Oracle (If you really like to pay)? What does SQL Server do that you can't find
The one that is capable of reverting to Paradox does indeed use the BDE (about the only thing that uses Paradox). I happen to know quite a bit about the BDE having worked on Builder/Delphi projects in the past. If the queries are done properly, it will work in either server or client mode, but you can always try the other. I loaded up the BDE administrator, and I managed to find some set of configuration that made it a little faster, but the queries ran both server and client just fine. The sad part is, I had better results executing on client side than on server side. Running select(*) on a table should never take very long, and most of the problem, that I can tell from diagnostics, was the server just halting network traffic. It may be a bug or problem with NT4 and SQL Server, but I'm sure not going to tell them they need to spend 20K to get these applications running. I'ld much rather develop a work-alike and drive the idiots that chose SQL Server in the first place out of the market if I can. Their product doesn't work for a group of people greater than 3 despite their claims. We've sank enough debugging time into that whole thing that we could already have developed an Alpha.
It's not the rogue master browser problem, BTW. Get out any NT4 system and any operating system built on NT since, and by default, it will be hideously slow. Pull out a Windows 98 machine and put in on the same network, and it will be instant. I'm fairly certain it's intentional incompatibility that they do all the time. Running mixed iterations of MS software will cause a lot of problems. They want you to upgrade everything all the time, because that's how they squeeze more money out of you.
I'm targetting any RDBMS with a JDBC driver, not just DB2. DB2 is just a preferred upgrade path. Even with this kind of portability, I don't take too much of a performance hit. Stored procedures wouldn't help, and neither would sub-selects. The only thing to gain from upgrading beyond MySQL for this application is good realiable data partitioning/clustering. No matter what anyone tells you, don't believe them that the current hot-backup crap in MySQL is useful for anything but failover. There are no distributed transactions the last time I checked, so I can't imagine that it would be reliable if you were using transactions on two different servers simultaneously.
If you are using Java, and you want to have a good abstracted database access, you should take a look at the open source hibernate project. Excellent software, much better than CMP. It supports all kinds of automatic relationship management too. It has it's own abstracted query language that it compiles into native queries for whatever features your DBMS can handle. It's the best way to get performance out of RDBMS's without spending months developing special queries for each DBMS, and still have DBMS portability.
Alright, it's time to bring out the big clue bat, because I've heard all this before, and I've tried it, and it's just bad.
.Net.
.Net isn't set up in a way such that my application will automatically run on a cluster. So, I can't sell it to small nor large businesses because it's a more expensive and less scalable system. See previous slashdot stories for other tales of the woes of how poorly .Net performs.
.Net was the right path. I have a larger market for my software, and it scales. MS SQL might work for some small niche of uses, but it doesn't compare to the database independance I have with Java. I choose the
Every program a client has that uses any version of SQL Server needs constant fixing, and is incredibly slow for 7 users. I could give you a list of programs, but it's just about every program on the market. One of them has an option to use paradox database files, and we have reverted back to that because it is both faster and more stable to share the paradox file over SMB. Every week, we have to repair an Exchange database even though it has never been improperly shut down. It generally fixes it's problems, but it's slow for the 5 people that use it. There are intentional problems between NT4 and Win2K+ using NT4 as a file server with SMB. In win2k, we side stepped the problem because MS had forgot to cause problems in the IPX part of SMB, and it worked fast when we disabled TCP/IP capabilities of SMB. In XP, it seems they caught on to it, and we just switched to Samba which was much faster, plus we get to use XFS, which has never corrupted or lost a file of mine. (NTFS has eaten the entire FS to "fix" itself in the past.)
That concludes all the problems I have with MS software, other than the fact they actually paid for the system that wasn't working.
Now, lets move on to your short sighted software development views.
SQL Server is a lot closer to Access than it is to a real RDBMS server. MySQL supports transactions, and has for a long time. If you don't need sub-selects, and a great majority of software doesn't, then you'll be just fine with it. I've had to use postgresql in the past as well, and other than vacuum, which can be cron'd, it's as fast as MySQL and it supports transactions. The only reason one would need to venture away from either of these two free options is if they needed real replication. Since I'm using JDBC, instead of some stupid MS recommended way of accessing the database, I can easily just change the configuration of the application and use DB2, the smartest database server available on the market. If you use one IBM prodect, like MS, you pretty much have to buy the whole lot (it's only just more economical with IBM, not a requirement like MS). My program, currently in JBoss and MySQL, can be deployed onto WebSphere and DB2 at any point. Too bad MS SQL isn't an option because replication has never worked.
I can sell a version of my same software to a small business down the street and they don't need to spend another dime on other software. I can sell to the little people that don't need the bloat of partitioned databases. In fact, I could sell it to work on one single machine, and it will run on windows just as well as Mac or Linux. They don't even have to buy a server. I can take my same program to Mr. Fortune500 and he can go buy everything he needs from IBM or SUN or BEA or Oracle, his choice, and I can set him up with the same program, except that it can handle 1000's of users, reliably, across a cluster.
Now, lets consider my alternative of having used
I now have a peice of software that requires $5K or more of other software in order to run. Now when I take it to a small business down the street, I have to charge 5K less, in MS tax, because they just want it to work. Now, lets say I want to take it to Mr. Fortune500. He can't choose anything other than MS.
There is absolutely no way you are going to convince me that
Actually, we supply a lot of small businesses in our area with whatever tech support they need. Kind of an outsourced IT staff. Paying us to fix things is as cheap as paying an MSCE monkey to spend 8 hours to fix a 5 minute job. We support OSS, so they save on licensing too. We even have a software team to make custom software, then release it open source.
.Net train before you can see where the tracks are going, then you go ahead. As for me, I plan to use as much cross-platform programming (mostly Java because the GUI is the same everywhere) and free/open source software that I possibly can, mostly because the products I use like JBoss (Free J2EE), Samba, MySQL/PostgreSQL/SAP/Firebird, etc. are more stable than .Net, Windows, MSSQL, etc.
The point is, they should be looking for the right service. You don't need dedicated staff with open source software. We get a call maybe once a month about an OSS product gone bad (usually something silly that can be fixed in 5 minutes if you know what you are doing), and we ssh in and fix it. We get calls about MS products and idiots that don't turn on things before they want to use them from 8AM till close every day. I'm pretty sure that most of our clients have spent more money on MS related tech support than OSS related tech support. I can calculate right now that the TCO for a pirated MS product would still be greater than a OS product by a significant factor. The speed at which MS products have to be fixed/patched is very much greater than a properly configured Linux system, and you're paying for that hell to boot.
If you want to shoot yourself in the foot by jumping on the
Before those of you that say the SQL Server is actually good start flaming me, that's where a lot of headaches come from. SQL Server drops records and corrupts more than MySQL before transaction support. (There, now I'll get flames from both ends.) Also consider the price you are paying. (Per connection last time I checked.) Spend more money on the hardware and get RAID-1 on good disks and a good UPS, and you will have a faster, more reliable RDBMS.
It's certainly got to be better than the bend over and take it like a machine method that this article seems to follow. Somehow I don't think a system is going to get much work done while it's being raped by some worm, if only for bandwidth reasons. Code signing is the only way to go for high security systems. Get a mathmatically verifiable hashing algorithm that would require an exponential algorithm to crack, and require multiple keys from multiple people in charge to actually sign an executable before the hardware would actually let it pass under the PC(program counter) register. That should prevent arbitrary code from every being executed. Now if you required signed data and connections, then you can spot breaches and roll back updates from those keys while still having a usable system assuming it's something that can be rolled back (not a nuclear weapon). If you have the right key to launch a nuclear weapon, then something had to have gone wrong on the real world side of things, where you will be shot for even thinking about stealing launch codes for nuclear arms.
Proper security can work. Sure there will be bugs in the software, but you better make sure that the hardware doesn't get all screwy, and you better guard the hardware physically as much as you do elsewise. If you are controling nuclear weapons, you better put as many armed guards on the box that controls the nuclear weapons as the sum of all armed guards that you have guarding each item it controls.
Best to implement two or more cleanroom implementations of the hardware so if one goofs, the other will kick in to actually have the computer "fall on the sword". It will be good to have them debug each other anyhow. Better be implemented on a tamper resistant CPU die so there is no cutting traces to avoid the security. Also you have to worry about the secrecy of both the private and public keys. It's only a matter of time before someone make a quantum computer that will give you a private key from a public of a symmetric key encryption algorithm. Yep, it would be a lot of hard work, but it's possible to build a computer that is gauranteed to be hacker proof to a calculable degree. The conclusion though would have to be that it would be easier to steel three keys to cause armageddon than actually look for a loophole.
Now that they've used it, I would think they would have known better :)
The answer to your question can easily be reduced to, "Do you have to support it?"
Seriously though, unless you hate the people that will be stuck with the mess, tell them it will take quite some time to get a solid version out. If they shoot themselves in the foot and tell you they need something by X, then when X comes and it's ratty and buggy and they point their fingers at you, pull out the logs/archives and point right back at them. When they want you to work extra to fix a bug, be sure to point out that they caused the problem, and now they will have to hire someone else to come in and help fix it. You shouldn't be penalized for their short-sited lack of management and planning skills. I would quit if anyone set a deadline earlier than I said was possible without hiring new help. My favorite is when you tell them you could have it by X done right, and they keep bothering you with other problems and wonder why you didn't have it done on time.
If you don't want to work 60h/week patching patches, I suggest you do it right to begin with. I work in a small little IT company, and people look to us to fix their software. A lot of the time we have to tell them their software was developed by a highschool drop-out monkey, and it would take less time to rewrite the whole thing than it would take to make their software work. Usually it's something running on Access written in VBA, that would take all of 40h to replace with a decent MVC serverside scalable extendable application using free or cheap software, assuming you know all the technologies involved.
241920000ms? 3 days? Never underestimate the latency of a station wagon full of tapes. Most of them can't make it to Wal-Mart down the street.
No, it's never the parents problem is it. If you want a crippled gun, get a BB gun. If you want a gun, and you don't want your children shooting each other, teach them gun safety, and LOCK the gun up, and lock the ammo in a different place. I haven't read a story about a kid being shot that didn't involve negligent parents. Why were the children in a room alone with a gun in the first place? If your children express interest in marksmanship, take them to the range and teach them gun safety. If you are too busy to take care of your children, you shouldn't have had children to begin with. But you were probably too negligent then too. Put them up for adoption. Children don't shoot themselves in the orphanage. If you're that negligent, they'll probably not notice a difference.
I'll try to point out the obvious flaws of your arguement without insulting you for not seeing just how obvious it is.
.Net will never fully work on another system without a legal battle. MS makes way too much money blackmailing people out of upgrades with the threats of no longer supporting their OS, thus leaving them open to disasterous bugs, rampant viruses, and crackers like they have NT4. If you have NT4 servers, you have been forced into paying thousands per machine because the product you bought was defective. In a decent sized company, that can hit millions fast.
There would be more jobs if competition existed.
Businesses could spend money on industries other than software if there was a reasonable alternative to Windows or at least competition to drive prices down. (They can't get their custom software to run on anything else because MS shops are the #1 ISVs because no one can get Dell, HP, Compaq, etc. to ship alternatives because of MS threats... which should be changing, but I haven't seen it.)
So, MS is on all the desktop workstations you can get without building them yourself. Why would anyone target a platform other than Windows? They really are locked in, and this unfair practice hurts the economy, not help it. My clients can't drop all their other software and just use mine, so it has to run on windows. No one has Linux, so it doesn't have to run there. Why go through the extra work of making software work with anything but Word, when that's the only thing anyone is going to have, and they can't change because they have invested millions of dollars in software that only runs on windows because there was no alternative to even develop for when most of it was developed.
The point is, MS is making more money than they should because they are a monopoly and they set the prices on their software, and people have to pay it, or pay the costs of having all their software ported to an alternative. This is MS's power, and this is why
MS doesn't deserve money because they refuse to fix holes in the software you bought. The item is defective, and has only marginally worked since it was released a decade ago, so naturally, you should have to also pay for the new product too! They are a drain on the rest of the world. Everytime you buy anything from a store, part of the overhead cost you are paying for goes to MS. How much do you pay in MS tax? If only they had to cope with the prospect of people just getting Linux and running their software on it without having to port millions of lines of VB and VC++ code, and retraining everyone, and converting all their data from MS formats to another format.
Oh well, too bad MS holding a gun to your data to get money out of you is actually good for the economy, eh?
Excuse my sarcasm, I know a lot of people see it your way, but that only makes me more angry with the situation. Effeciency will help the economy, not the blindness to corperate corruption that seems to be the staple of the Republican party lately.
Both you, kris, and anthony would probably do yourself well by taking a look at Jetty or Resin.
:) I have a hard time believing he was serious though, I woulda probably modded him as funny. It's sad someone actually fell for his post. There really is no speed or stability difference, it's just easier to make small sites with PHP, and large sites (or integrate with legacy systems) with J2EE.
Java doesn't leak away memory unless you are running in a Solaris machine, or someone coded a class that keeps pointers to objects that should have died long ago. If you don't handle state properly, and you keep it in memory but don't periodically check to see if it expired, of course you are going to have problems.
Some JSP engines store persistance on disk, in a database, or across a cluster in either variety as well.
When people quote Tomcat and JRun as reasons why PHP is faster than JSP, I just want to shoot them in the head. At least give Jetty (open source) a chance even if you don't give Resin ($$$), BEA ($$$), or Websphere ($$$) a chance. I wouldn't trust Macromedia (JRun) to make any server-side product really... Anyone that thinks you don't need a sine function in a language to display flashy graphics obviously doesn't have much going on in the head. Tomcat was a donated reference implementation. It's primary goal until recently seems to have been stability. Now they seem to have switched gears for performance. Jetty on the other hand is appearantly programmed by a savant. Within days of the release of NIO, he seems to have reworked the entire engine to support NIO. I really don't see how anyone could not be pleased with jetty though. I set it up on a dual processor server at the local university, and classes of 15 start and restart Jetty and they don't complain about speed at all. JBoss embeds Jetty because it's so good. Jetty doesn't support working through Apache (that's what makes Tomcat slow more than anything I would say), so you may not be happy with that, but just about any project worthy of J2EE is worthy of 100$/mo dedicated hosting.
Be sure, in any case, to try different VM's to see what works best for your project. The IBM JRE seems to be the fastest, but some argue that BEA's VM (JRockit) is actually faster for server-side stuff. Both should be free, but I'm not sure of the licensing terms.
I've seen some other FUD around going the other way too. PHP's Zend compiler is a JIT compiler as well. The speed of the two should be comperable. They both have their place, but I will probably never recommend PHP for large projects because it's so hard to seperate the logic from the presentation, where there are at least 3 tools to do this with Java.
If the future of the Web is rich clients, I'm curious if PHP+Javascript+XUL will turn out better than Java Webstart. It will be interesting to see what happens, but I'm banking on Java because of the massive numbers of backers in the industry.
I hope no one got the wrong idea, but I love PHP too, and am very excited about the new version. But that one crazy AC trying to compare Java and PHP the way he did was just silly
Because java forks processes, sometimes the OS will lie about how much memory Java is actually using. Turn on the Thread count option so you can see how many threads the process has. I can run Eclipse with -Xmx4M but not 2M. When I run it with 4M, I get 30M usage in 11 threads.. 11*4=44M which would be the maximum windows could report.
If you have 11 threads, and you are using 300M, you are getting around 90% usage of a 32M heap. This means the VM has to relocate memory in order to bring up new windows, and text files you have open will be fragmented and/or garbage collected when you switch away.
Really, you should try -Xmx256M, and I bet it runs much smoother.
The default settings only lets java claim 64M of RAM last time I checked.
Isn't it funny that everyone complains about Java's memory footprint when in fact the memory usage is usually about 10% of what they think it is because Java spawns at least 4 threads, and usually 5-6. Console applications use about 5 threads and consume about 1M of RAM, but they get reported as 5M or so RAM.
By default, the OS will report each individual thread as having access to the heap, so each thread is claiming about 3M of RAM in eclipse on my computer. So, windows adds them all together (11 threads) and tells me there is 30M of RAM in use. (It will actually start with only 3M of heap on my computer, but it's slow.)
People on Linux systems add together all the VM memory usage and say Java is using 1G of RAM on my linux machine (because Java on Linux uses more RAM for some reason.) The real way to calculate the ammount of RAM in use is to take the RSS field of "ps" of the largest process (they should all be the same, but sometimes they have some handles that are just allocated to a single thread where the heap is accessable to all.).
Actually... I was running Eclipse on a Pentium 166 with 32M of RAM once. It was horribly slow because it spent most of the time swapping memory to disk. When the hard drive light wasn't constantly on, it was usable, but that situation rarely happens when X takes up all the memory already.
You can't compare versions of software that way anyhow. Software usually gets slower every revision just because. If it wasn't that way, then Windows XP would run really fast on a Pentium 166 since it is 5 generations of software more than Win95.
I don't know what you are doing to make Eclipse slow on your machine with 512M of RAM. Maybe the Webshere version just has so many plugins that you actually should set the heap size up. Are you running the latest JRE (IBM doesn't have a 1.4 for Windows the last time I checked)? I would definately say there is something wrong though if Eclipse is slow on your machine. Check how much memory it is using. If it's close to 64m, then it has filled the default maximum heap size.
Try this:
java -Xmx256M -cp startup.jar org.eclipse.core.launcher.Main
You may want to play with the 256M number since you have a lot of RAM. I hope SUN comes up with a way to automatically optimize the maximum heap size dynamically.
Either way, languages are all derivatives of something else. We aren't grunting over a peice of meat, therefore keeping a language pure is idiocy to an even greater extent.