Slashdot Mirror

The Ars Digita story is a classic example of what Cringely is talking about- a company run into the ground by "professional" managers brought in by the VCs. Here's the story, as told by one of the company's founders:

http://eveander.com/arsdigita-history

Even though it's "just one side of the story," the consensus is that it's pretty close to what really happened.

In the end, the VCs cut a deal with Redhat, who hired a few of Ars' staff to make it look like the company was successfully sold. Fortunately, Ars' great products live on as open source software, OpenACS, and Redhat's CCM. Though Ars' incompetent management pushed CCM as the next, great version of their software, it was never more than vaporware. Redhat has continued to develop it, but it's still not finished.

For anyone interested, this months Linux Journal has an article on OpenACS (page 12) which reads, in part:

"It's easy to say that OpenACS is a toolkit for creating on-line communities. But what does that mean? For starters, it means that OpenACS comes with working versions of most of the applications you're likely to want on a community web site. It handles user registration and administration, forums, FAQs, group (including a rich permission scheme), news updates, file storage and distribution, personal home pages, surveys and a we-based calendar. As you might expect from a modern system, administration of the application is done almost completely through the Web, with only a few configuration files."

From OpenACS' frontpage:

WHAT is OpenACS?
OpenACS (Open Architecture Community System) is an advanced toolkit for building scalable, community-oriented web applications. If you're thinking of building an enterprise-level web application, OpenACS is a solid, proven foundation that will give you a 3-6 month headstart.

I would not fault MySQL for this, though, since after all it was designed and still mainly used as an SQL wrapper for flat file data, and this is why it's usually much faster than full-featured RDBMSes. The problem is with mindless open-source advocates who try to pump up MySQL as the be-all, end-all database solution. For a personal website or small business, MySQL is more than adequate, but its lack of higher-end SQL features make it a poor fit for large, distributed, mission-critical corporate or university data storage.

IBM and Microsoft's customers are generally in this higher end of the database spectrum, where Oracle or DB2 makes much more sense. It's no surprise that they would want to put MySQL in its place as an entry-level database system, where it belongs, and I fail to see how this story qualifies as news.

It seems your question is a bit badly researched before "Ask Slashdot", but anyhow..

I have personal experience with the following :

- eZ Publish (PHP, *nix, MySQL/PgSQL based)
- Slashcode (Perl, *nix, MySQL based)
- PHP Nuke (PHP, MySQL based)

I wouldnt recommend PHP Nuke, but have little experience with the forks off it.. (Check Freshmeat.net as someone suggested).

eZ Publish is the closest I have seen a complete GPL CMS (Content Management Solution), and integrating some of the addons or buying the desktop edition makes it very easy to use!Highly recommended, and now comes in an easy to install Debian package too ! :) (As of woody, apt-cache search ezpublish and then apt-get install the package ..)

Slashcode is possibly one of the better weblogs, although you should possibly check these too, if that is the kind of website you need :

- Scoop
- Drupal

For non "LAMP", based on Tcl and the AOLServer webserver, check out OpenACS, which is reportedly very feature rich.

I do not have personal experience with either Scoop, Drupal or OpenACS, but several sites use them and produce great sites with them.

Good luck !

Another option is OpenACS. Lots of neat tools for building dynamic websites.

Intranet module of above has projetcs/tasks/tickets functionality all with calendars & billing, time sheets and history tracking, user comments and customer/partner access, automatic mailing of reminders.

Requires AOLServer + Oracle or PostgreSQL. Free in all senses -- http://openacs.org/.

Gant charts are not there, though :^)

It didn't have transactions, and the only way to get them now is to use InnoDB (this is NOT used by default), which isn't completely integrated. InnoDB and BDB are taken from other projects, and would be better on their own.

MySQL developers have made claims that transactions and rollbacks are a bad thing! I kid you not.

They claimed one could have atomicity without rollback. Okay, so what do you do if a SQL statement which is part of a (user emulated) "transaction" fails. You are stuck halfway through, or else you might have to do a SQL statement that undoes what you did (good luck). And if that SQL statement fails, you are hosed.

MySQL does not (by default) support Atomicity, Consistancy, Isolation and Durability (ACID).

Their developers appear to not know the meaning of these terms.

MySQL isn't truly open source.

MySQL isn't a real database, it is a SQL interface to a file system. MySQL isn't much better than using flat files, and due to the complexity, is often worse.

PostgreSQL has none of these problems. And the performance is much better than the old versions.

Only use MySQL if you don't care about your data. Yeah, it might be fine for a web counter where if it gets hosed, big deal.

If you care about your data and need a REAL database which is actually Open Source, use PostgreSQL.

See this article:

Why Not MySQL.

To quote the release.... "The application is a service for tourists visiting the historic city of Bath (in the UK). It guides users around the city and towards major attractions. At the same time they are involved in a 'Treasure Hunt' answering questions related to historic sites, requiring users to visit them. Prizes can be won and sponsorhsip opportunities exist for businesses located on the trail."

If I recall correctly, the backend tying openACS to the SMS gateway has also been opensourced. Check it out.

To quote the release.... "The application is a service for tourists visiting the historic city of Bath (in the UK). It guides users around the city and towards major attractions. At the same time they are involved in a 'Treasure Hunt' answering questions related to historic sites, requiring users to visit them. Prizes can be won and sponsorhsip opportunities exist for businesses located on the trail."

If I recall correctly, the backend tying openACS to the SMS gateway has also been opensourced. Check it out.

Look at openACS. Its a fast evolving toolkit, with a lot of features out of the box. The current project website is not the best looking, but the toolkit has been used to develop a lot of interesting sites.

Methodology
If I were you, I would stay away from the Vignettes and other off-the-shelf CMSs. To paraphrase Phil Greenspun, these guys pricing works along the lines of ... shake the customer by his feet and see how much money falls out, then charge another $50 000k for support.

I would also not be in a rush to implement a totally custom solution. Building from scratch is usually a dumb idea(tm). No point in reinventing the wheel. Having said that there is a slight difference between a Michelin-clad Ferrari wheel and a 0BC Roman chariot's wheel...

I agree with you, do not go for the slashdot look. That virtually rules out most of the nukes (phpnuke, postnuke, drupal, slashcode etc). It is

so boring

so overused

suitable for weblogs and news sites but not for more mainstream content sites.

oss is good
The beauty of using OSS toolkits is that you get a head start. If any consultant (read salesman) tells you that their product fits your needs perfectly, then a. shoot them, b. chop them into little pieces c.feed them to the snakes d. shoot the snake ..... just for good measure.

The best that you can hope for is to have a basic and solid foundation that you can build on.

decisions
Some of the things to look for include the following:

• ability to handle workflow.
• ability to deal with permissions
• ability to deal with authentication
• ability to handle more than just plain text
• ability to version ... rollbacks ... track changes
• ability to handle templates, ... proper templates, not just color changes.
• level of developer support
• level of developer competence
• pace of changes

For each toolkit, look at sites that have implemented it. If they *all* look the same, steer clear. Its a sure sign that templating was poorly implemented, or that the toolkit is difficult to customise.

Post a couple of questions on the boards. If the tone is friendly, then you know that if you did pay these guys to do work for you, the service would be great.

If you are building a proper CMS, its going to be painful.

and you win an all expenses paid tour of some of the sites built using openACS and its cousin ACS classic.

• Grand Expeditions
• Wine Access
• photo.net
• Stirling Engines
• Online Ireland and more here
• Away
• Sherman power sports
• Net effect

Look at openACS. Its a fast evolving toolkit, with a lot of features out of the box. The current project website is not the best looking, but the toolkit has been used to develop a lot of interesting sites.

Methodology
If I were you, I would stay away from the Vignettes and other off-the-shelf CMSs. To paraphrase Phil Greenspun, these guys pricing works along the lines of ... shake the customer by his feet and see how much money falls out, then charge another $50 000k for support.

I would also not be in a rush to implement a totally custom solution. Building from scratch is usually a dumb idea(tm). No point in reinventing the wheel. Having said that there is a slight difference between a Michelin-clad Ferrari wheel and a 0BC Roman chariot's wheel...

I agree with you, do not go for the slashdot look. That virtually rules out most of the nukes (phpnuke, postnuke, drupal, slashcode etc). It is

so boring

so overused

suitable for weblogs and news sites but not for more mainstream content sites.

oss is good
The beauty of using OSS toolkits is that you get a head start. If any consultant (read salesman) tells you that their product fits your needs perfectly, then a. shoot them, b. chop them into little pieces c.feed them to the snakes d. shoot the snake ..... just for good measure.

The best that you can hope for is to have a basic and solid foundation that you can build on.

decisions
Some of the things to look for include the following:

• ability to handle workflow.
• ability to deal with permissions
• ability to deal with authentication
• ability to handle more than just plain text
• ability to version ... rollbacks ... track changes
• ability to handle templates, ... proper templates, not just color changes.
• level of developer support
• level of developer competence
• pace of changes

For each toolkit, look at sites that have implemented it. If they *all* look the same, steer clear. Its a sure sign that templating was poorly implemented, or that the toolkit is difficult to customise.

Post a couple of questions on the boards. If the tone is friendly, then you know that if you did pay these guys to do work for you, the service would be great.

If you are building a proper CMS, its going to be painful.

and you win an all expenses paid tour of some of the sites built using openACS and its cousin ACS classic.

• Grand Expeditions
• Wine Access
• photo.net
• Stirling Engines
• Online Ireland and more here
• Away
• Sherman power sports
• Net effect

The best thing about the openACS toolkit is that you can have a functional db-backed site up in no time. No need to re-invent the wheel. :)

I'm sure Zope is much simpler than it looks, and after working with it awhile, you get to a point where it's all crystal clear. But I could never get to that point- everything is obscured by silly new paradigms, acronyms, and taxonomy. It's all just too zilly for me. After ztruggling with Zope for what zeemed like a zentury, I picked up OpenACS and built a website with it in about three weeks.

Hejsan Lasse! :)

Disclaimer: I have used MySQL (a little) and PgSQL (a lot more) but not FB (nor IB), so I'm just repeating what I have read somewhere.

My impression is that FB and PgSQL are quite similar. Both have multiversion cuncurrency control, ACID compliancy, stored procedures, triggers, referential integrity and a lot of other cool stuff. Both support large subsets of SQL92, outer joins, subselects etc.
PgSQL, on one hand, is a classical unix style forking multi-process server, whereas FB has both a multi-process and a multithreaded model. The advantage of the multithreaded model is of course a shared cache, but FB:s threaded version doesn't work on multiprocessor machines. Anyway, on unix at least PgSQL uses posix (or was it sysv?) shm to get a shared cache. On the other hand, the windows version of PgSQL is a cygwin hack, whereas FB has a native win version.

MySQL, on the other hand, does not (currently) play in the same league. Yes, with simple queries and low concurrency (albeit a common situation) it probably smokes both FB and PgSQL speedwise. IMHO MySQL suffers from limited SQL support, table level locking and lack of ACID properties. For more info check out this page which describes some of the problems with MySQL in more detail (and of course there is a super-long flamewar at the end too...). Anyway, MySQL 4 (alpha) apparently supports more SQL constructs, and new tablehandlers give it row-level locking (or even multiversion concurrency control) and at least some limited ACID capabilities, so the gap between MySQL and PgSQL/FB appears to be shrinking fast..

And then there is of course SAPDB, which appears to be very impressive according to the few people who have tried it.

In the end, I think all of them are excellent (with some reservations towards MySQL and SAPDB). And of course they are all free (duh!). I'd say FB has better windows support than PgSQL and certainly has better delphi drivers, whereas PgSQL would be better on a multiprocessor unix/linux box.

There is discussion and experimentation with ways to add Java to the mix for those who need it. See the current thread about related AOLserver modules ns_java / ns_javablend and their current status, on the OpenACS.Org site.

It's not especially relevant to the Java question, but I'm the person who tries to create Red Hat RPMs for OpenACS.

There is discussion and experimentation with ways to add Java to the mix for those who need it. See the current thread about related AOLserver modules ns_java / ns_javablend and their current status, on the OpenACS.Org site.

It's not especially relevant to the Java question, but I'm the person who tries to create Red Hat RPMs for OpenACS.

CMS is software to define rules to manage content production and publication. So if you were to download and install the aD CMS, you would get all of the above, except integrated. So you post content as with the publishing system like Slash, but you can collaboratively author and manage said content a la SourceForge. The previous four are part of the basic ACS system; they are necessary but not sufficient to describe CMS. Additionally, the focus of CMS is to manage content -- so the CMS software also allows you to write the control flow of content and integrate it with all the above. So rather than being limited to the rules for posting on /. or Scoop, you can define the rules for "such and such must review, approve here, loop and edit, comment, publish, email, repeat" or whatever you come up with on-site.

Of course, unless you have a penchant for Java-flavoured pain, it might be easier to use the CMS with OpenACS 4.5beta1 than the packages from a defunct company that fired most of their programmers. Still, it's nicer than what Vignette will charge you 6 figures for.

Not only that, but solve as little of the problem as possible. Someone else has probably already solved much of your problem and made available library/framework code that you can build on. The perfect language isn't if you have to re-invent every wheel with it. I do try to follow my own advice -- the availability of a partial solution for one of my problem domains is why I use Tcl a fair bit despite hating the language. OTOH if you want a language with cool features for the sake of coolness, I'd go with O'Caml.

Greenspun may not have been the most pleasant boss (I have no experience of this personally) or the wisest possible businessman, but compared to the VCs and managers who took over the company he looks like Solomon. It's hard to credit the amount of sheer stupidity in the running of the company after the takeover. It must have taken some real effort to take a profitable, slow-growing company and turn it into a loss-making whale which didn't have any non-vapourware products.

I think that this is a software equivalent of the Edsel story: be very wary of changing product direction solely for marketing and not technical reasons, to give customers 'what they say they want' (in this case Java). And don't let the engineers get sidetracked into building something horribly overengineered and way too complex (second system syndrome).

I agree that Philip and co.'s code is crufty and difficult to maintain: but it's an absolute dream compared to the never-finished successor in Java. FWIW, the Tcl-based ACS is still being maintained as OpenACS, they have a port to Postgres (as well as Oracle) and the forthcoming version 4.0 is progressing nicely.

They made the ACS(ArsDigita Community System) and offered programming for it. It is currently written in java and runs under Aolserver. What is offers is:

• message boards
• user management
• Ecommerce
• Content management - that means having a whole system of library articles that can be edited through a web interface. That is the overlysimplified answer.
• Intranet type functions
• Ticket tracker
• Much Much More

Luckally OpenACS exists for future ACS incarnations that do NOT use oracle for its database.

So to answer no, it is not rsync and it is not for people that are "stupid." It is actually a lot of useful code. My guess is that RedHat is going to try and sell a website in a box.

"The terms of the settlement were not disclosed, but Greenspun has since purchased an RV and an airplane". So at least we know he did OK! Beats me why RH bought the company, when they could have brought in - openacs

ArsDigita switched to the ArsDigita Public Liscense a few months ago when they released ACS 4.6. As you can imagine, this pissed off many people in the community. However the folks at OpenACS have ported ACS 3 and ACS 4.2 (both under the GPL) to Postgresql. Work continues, unabated by short-sighted VC's.

ArsDigita switched to the ArsDigita Public Liscense a few months ago when they released ACS 4.6. As you can imagine, this pissed off many people in the community. However the folks at OpenACS have ported ACS 3 and ACS 4.2 (both under the GPL) to Postgresql. Work continues, unabated by short-sighted VC's.

See also this thread on OpenACS bboards for more info, and also Philip Greenspun's comments

Twas a good thing - just remember, however greedy you get, never succomb to the temptation of VC

See also this thread on OpenACS bboards for more info, and also Philip Greenspun's comments

Twas a good thing - just remember, however greedy you get, never succomb to the temptation of VC

Yesterday evening, after I read about the new iMac and in particular iPhoto, I showed the new stuff to my wife, carefully guiding here to agree that we might buy one of these slim new machines. We already own a PBG4, an older PPC 96000 apart from Wintel machines running W98 or Suse Linux. So far, my wife hesitated to touch the Macs since she feels technologically challenged (like many people) and is happy working with M$ Word and Internet Explorer on the PC.

I tried to argue how easily she could herself create digital albums of our 5-month old son, or create video clips using iMovie together with a digital videocamera (to be bought too). So far, I've been afraid to ask her to handle the publishing process starting with importing the JPEGS from our memory card, using Photoshop to edit the image (e.g., eliminate red eye effect), crop or resize the images, and finally print or publish the images.

Without boring you, the bottom line of my message is: attracted by the excellent hardware and software design of the new iMac, my wife got really enthusiastic about learning how to manage digital content with iPhoto and perhaps iMovie. I am very glad about that fact because this frees me from publishing and printing every single photo on my own, a spend more time on building up the homepage of Jannik and learing Zope or OpenACS.

Check out OpenACS, continuing development of ArsDigita's Arsgitia Community System. They've had discussion, commenting and other community/collaboration features for years and there are people working on adding more wiki and weblog-like functionality.

There's a good article here called "Why Not MySQL?" by Ben Adida [mailto], part of the OpenACS Project [openacs.org], on why MySQL wasn't the right choice for OpenACS (at the time). It's quite out of date (and is recognised as such by the author), but still worth a read, and there many interesting submitted comments. Take a look at some decent free RDBMS alternatives such as Firebird (open-source free Interbase) or PostgreSQL while you're at it. Oh, and there's plenty more dicussion on MySQL in a previous Slashdot article here.

Stef

There's a good article here called "Why Not MySQL?" by Ben Adida [mailto], part of the OpenACS Project [openacs.org], on why MySQL wasn't the right choice for OpenACS (at the time). It's quite out of date (and is recognised as such by the author), but still worth a read, and there many interesting submitted comments. Take a look at some decent free RDBMS alternatives such as Firebird (open-source free Interbase) or PostgreSQL while you're at it. Oh, and there's plenty more dicussion on MySQL in a previous Slashdot article here.

Stef

called "Why Not MySQL?" by Ben Adida, part of the OpenACS Project, on why MySQL wasn't the right choice for OpenACS. It's quite out of date (and is recognised as such by the author), but still worth a read, and there many interesting submitted comments.

Get it here

Not that anyone tends to read/moderate Slashdot posts after they're a day old, so few will see this...

called "Why Not MySQL?" by Ben Adida, part of the OpenACS Project, on why MySQL wasn't the right choice for OpenACS. It's quite out of date (and is recognised as such by the author), but still worth a read, and there many interesting submitted comments.

Get it here

Not that anyone tends to read/moderate Slashdot posts after they're a day old, so few will see this...

The above is not a troll, it is the truth. MySQL is junk compared with PostgreSQL. Why anyone would use it is beyond me.

Another way is to do something for the community, which everyone would love. For instance, help purchase and colocate a server for the community to develop on. Or host a social for the developers to meet and drink beer together. Or help purchase tshirts for everyone to get.

The best way, IMHO, to show support for an OSS community is to help keep it live and vibrant.

talli

I use @home, but haven't been effected yet. I know because I haven't recieved an email alert from my offsite server monitor (I also double checked by logging into a remote server via ssh and testing the server with lynx).

Check out http://uptime.openacs.org/uptime/ for the free (and open source) monitoring service.

Never use MySQL to store anything important. Use PostgreSQL, InterBase, or some other Open Source DBMS which passes the ACID test.

Check here for a discussion of the problems with MySQL.

The difference, though, is that MySQL does not pretend to be a real DBMS. It's problems are well known and documented. Even Las Alamos couldn't mistake it for one. MS SQL, on the other hand, is marketed as a enterprise level RDBMS. Consequently, its users have an expectation of a resonable level of quality.

In short, it is acceptable for MySQL to lose data, it is not acceptable for PostgreSQL, Oracle, Interbase, MS SQL, or any other database in their league to behave in a similar manner.

• My friend is using it, and he's happy with it.
• My ISP only has MySQL support, so it's the only way to go.
• Even Slashdot is using it!
• It's the most popular DBMS, so it must be the best, right?
• Everyone and my neighbour is using it, so it's easy to get support.

All the best rumors say so...

RedHat to BUY GreatBridge (postgreSQL) ???

If they are planning to compete with Oracle, what other GPL database is there?

Sap DB has many features and administrative tools still lacking in PostgreSQL.

Try real archiving and replication, for instance. Being able to specify where to allocate tables and indices without using "ln -s". An overwriting storage manager so you don't need to VACUUM nightly (or hourly as some do on very busy systems).

That just scratches the surface.

Yes, PG is vastly improved and I'm a big fan. And much of the above list is scheduled for PG 7.2. But today it's PG 7.1 and PG doesn't have these features.

OpenACScurrently is supporting Oracle and PostgreSQL with our upcoming OpenACS 4.x product. Sap DB is very likely to become the third RDBMS we support with the toolkit.

If you can stomach tcl, it's an interesting tool for promoting online collaboration.
(disclaimer -- i work closely with the project coordinators.. etc..)

since no one has mentioned it I thought I'd dig up a few links I remember reading.

• From the PostgreSQL FAQ: "What is the maximum size for a row, table, database?"
• semi-old article on why not to use mySQL
• a comparison of mSQL, mySQL, and PostgeSQL: part 1, part 2, part 3
• and if your lazy: google search for "linux sql database"

From the PostgreSQL FAQ, linked above:

• Maximum size for a database? unlimited (60GB databases exist)
• Maximum size for a table? 16 TB
• Maximum size for a row? unlimited in 7.1 and later
• Maximum size for a field? 1GB in 7.1 and later
• Maximum number of rows in a table? unlimited
• Maximum number of columns in a table? 250-1600 depending on column types
• Maximum number of indexes on a table? unlimited

If you couldn't tell I like postgres but as a business you should get what you think is best. From what I've heard Oracle on Solaris is where it's at if you can afford it. (I can't)

Leknor

OpenACS (Open ArsDigita Community System) is an advanced toolkit for building scalable, community-oriented web applications.

[shameless plug]: the openacs. Uses PostgreSQL, can handle high-volume.

Did you consider PostgreSQL?. DB2's a fine database, I'm sure, but from your post it doesn't seem you are aware of alternatives to MySQL in the Open Source world... PostgreSQL 7 is a lot more fully featured than MySQL, and 7 is much faster than previous releases... faster than MySQL if certain benchmarks are to be believed. (I do not know, however I could understand if it scales better than MySQL...) For a little more info: Why OpenACS does not use MySQL.

Note: above post is not meant as a flame towards MySQL, I am just trying to point out possible alternatives which may have been overlooked. Thank you.

I'm surprised you are having so much trouble with Mozilla on a 650MHz Athlon. I run it on a 466MHz Celeron, 64MB RAM, and although it takes a bit more time than Netscape to load, its not unusuable.

1. PostgreSQL does run under NT. It runs under cygwin, and I have no personal experience running it under NT, but it does run under NT (and, yes, I do mean the server, not just various clients).

2. In PostgreSQL, writers never block readers, so your statement to the contrary is a factual error. InterBase and PostgreSQL use a very similar paradigm for the storing of actual data, and InterBase's "Multi-Generational" stuff is equivalent to PostgreSQL's "Multi-Variate Concurrency Control", or "MVCC" for short).

Again, I'm not disparaging your use of InterBase, a fine product that hopefully will survive the current political machinations that surround it. But I can't let your misconceptions about PostgreSQL pass without comment.

The methods are a new DB API function db_dfs (depth-first search); and an older tcl-layer fxn which I am unfamiliar with, but I know of its existence. Ask at the above.

Philip Greenspun's book (reviewed/interviewed on /.), Philip and Alex's Guide to Web Publishing (full text free online) has some interesting coverage of this stuff, too.

While Imedix is a cool product, I don't think it will meet his needs for a general purpose EMR.

The ACS (or OpenACS) can be customized to suit a variety of tasks, but no EMR module yet exists. So be prepared to do a little bit of hacking/data model extension. Drop me mail if you are interested in taking this route.

A better bet for what you're trying to do (at this point) is probably GNUmed. It uses Postgresql for its database layer (passes the ACID test). Dr. Horst Herb and his crew have built the software in a very thoughtful fashion -- they've clearly used lots of commercial products before.

One final thing to remember for those who wish to go it alone & code their own product is the importance of security.

Good security is critical for all EMR software, but is of particular concern in the USA, where HIPAA rules are starting to be implemented.

Because of these rules, the task of developing an EMR for use as a permanent medical record has become more time consuming, and complex. The law is, ultimately, a consumer protection act, and should lead to higher-quality, standards-driven EMRs in the future. Be sure to examine the rules before setting out on any healthcare project.

hope it helps.

...running long multiple transactions where you get a conflict at the end (in this page/row locking is better)

We here at MySQL has always tried to design very fair test that no one can misinterpret or lie about.

The way to set up the databases in a test is also very crucial for the performance of the database. The article doesn't mention anything about this or even with which ODBC driver they used the different databases

MySQL also have two ODBC drivers, one slow with debugging and one fast. It should be very nice to know how they actually did use MySQL. To get any performance from Oracle, on has also to tune this a lot; The ODBC driver for Oracle has also very bad performance; This is a common known fact; No one runs a critical system with Oracle and ODBC.

If the ODBC driver were the cause of MySQL's degradation under low, then the tests with (presumably) Oracle would presumably also show it degradating similarly under load. Which, of course, they don't. The "real" RDMBS systems scale reasonably well, while MySQL performance drops quickly as the number of simultaneous users increases.

As to why MySQL perhaps performs poorly in these tests under light load, another explanation might well be that even MySQL's benchmarks show Postgres beating MySQL on a JOIN test. Most real-world work with databases involve queries which JOIN tables, and the real-world benchmark suite used very likely penalizes a database engine which performs poorly on JOINs. As is the case with MySQL.

Of course, the "fair" MySQL tests on their benchmark tests are single-user tests. This doesn't represent the running environment most folks putting up web pages aspire to. When I put up a web site, I'm hoping for many simultaneous users, not just one. MySQL's "fair" tests hide the fact that MySQL falls apart under load by "fairly" not including tests with multiple users.

One thing that also is interesting is that they don't mention which PostgreSQL version they are using. It's very unlikely that they did actually test PostgreSQL 7.0 as this has at least one very fatal bug in the index handling which made it useless for benchmarks (at least when we did a test run on it).

I hope Monty had the common courtesy to submit a bug report on this "fatal bug" they ran into.

Threads also gives MySQL better scalability than processes, that PostgreSQL uses, so we are very confident about the future.

It will speed start-up overhead, but anyone in their right mind using an RDBMS in a web environment uses persistent, pooled connections anyway and start-up time is in practice not an issue. The AOLserver database API supports persisten pooled connections in C, Tcl, Java and Python (the latter two are in pre-release form) in a database-agnostic way, so it's particularly not an issue if you choose your webserver intellegently.

The net result is that the posted test is about as wrong as you can do a test, the important thing is just to get the people that reads that page to understand that.

And $Oz2500 PER CPU for an "Internet Connector" licence if we want our SQL database to be web-enabled.

If you think that's absurd, look at what Philip Greenspun has to say about RDBMS pricing. M$ $QL $erver is actually one of your cheapest options, and one of only two commercial RDBMSes that actually publishes prices. (SOLID is the other one.) M$ $QL $erver, for what it is, actually isn't that bad (they stole most of it from Sybase), although it bites big for most web stuff because it supports a laughable 256 character VARCHAR and because all of the JDBC drivers are closed-source and expensive (you can use the ODBC-JDBC bridge, but it's slow).

Of course, I'd recommend PostgreSQL if you don't want to pay for a database, or MySQL if you don't want to pay for a filesystem. :-)


~wog

Then there's the aolserver, whatever the advantages of that over apache I haven't a clue.

Gotta stand up for my webserver of choice, here :)

AOLserver is a full multi-threaded webserver that has a built-in TCL interpreter. It's easy to install, and the latest version (3.0) is wicked fast. It also excels at database connectivity.

Due to it's threaded nature, and the pooling of DB connections, it tends to be faster than Apache in dynamically generated web sites. An Apache God could probably match AOLserver with careful tuning, so speed isn't really much of an issue.

I just like AOLserver better than Apache, mainly due to the OpenACS guys and the work they've done. Great out-of-the-box functionality (and, soon available for Apache as well).

AOLserver 3.0 is also open-sourced (a true GPL, I believe). Be nice to those guys -- they did the right thing with AOLserver; maybe they'll do the same for AIM (errmmm.... maybe).