Do remember that fire safety codes typically require a data center to cut all power, including all UPS power, if the EPO switch is used. That should take down any single-location site which is following those codes. It's not the database server but the fire requlations which mandate this behavior.
Options include ignoring the law, operating in places with different law and using multiple locations.
The problem in two loss of power incidents, LiveJournal (EPO) and Wikipedia (two circuit breakers tripping) was disk controllers of two brands, both with battery backed up RAM, which didn't turn off the drive write buffers and lost data in spite of the battery backup. One of the vendors used to do that and removed it. It is quite helpful if the disk system doesn't throw away data it claims to have securely written.
Both places had and have a one data center vulnerability and can't really expect five nines reliability, since that's above what a single tier IV data center can be expected to assure - limit is 99.995%. Both are planning to move away from that single site vulnerability. Wikipedia is still in two to three nines territory, and in a colo which is providing tier II-III capability, and likely to stay that way for two or more years.
Both probably know what it'll take to get five nines in system architecture. Wikipedia, me included, certainly does, but isn't currently trying. It's still using donations mainly to keep up with growth on a high cost-effectiveness path, not a very high reliability path. OK enough for its route to the top few sites on the net funded by donations but you wouldn't want to do it with so little funding if downtime or less than ideal performance cost money.
The MySQL libraries are not exclusively GPL licensed, so it's not mandatory that an application using them be GPL licensed. A list of 20 other open source licenses which are accepted by MySQL for use with its client libraries can be found at http://www.mysql.com/company/legal/licensing/foss- exception.html . The PHP license is one of them.
MySQL does have two employees in Brazil, one the manager of the group which does most of the bugs verification work. Not the same as the whole company but good anyway.
Just the usual anti-MySQL FUD. If you're interested in transactions you'll probably be using the standard InnoDB storage engine and that not only supports the usual transaction isolation levels, it also supports the XA Distributed Transaction standard. You can find some worked examples of transaction isolation levels in the blog entries at:
I disagree about the significan attack vector for vandalism, because as you said "these pages can remain vandalized for months until someone comes along and finds the page". If nobody is finding it, it's insignificant because nobody is being harmed by what they aren't seeing. It does take a degree of understanding of the nature of the work and its building process, and the self-restraint and judgment not to over-react when finding something obviously ridiculous.
It is good to try to improve quality assurance efforts for such things, though.
The key reason given for the "experiment" of no anonymous page creation was a lack of inward links to such articles. One alternative approach would have been to specifically target that: prevent new page creation only for unlinked articles. That would more specifically target the expressed problem and reduce the large side-effect of a total block.
A one third reduction in the rate of new article creation isn't a slight reduction. Remember that the expressed goal of the project is to contain the sum of human knowledge. A million pages is a tiny proportion of that - we're still in very early days of the project's growth and there's a tremendous amount of specialised and even quite general knowledge which isn't present yet.
So far in all the reports of his words I've seen, he's reported that the people he mentioned it to didn't believe it. Makes it pretty unlikely he was humiliated.
Agreed that libel per se as in murder is still libel, but that's not what happened here - the closest the offensive content came was suggesting an investigation, which both of us know would have applied to everyone with even passing connection to Kennedy, such was scope of the many investigations of the matter.
A wiki of this sort has no choice but to not use prior review. Its core principle is that anyone can contribute and anyone with significant knowledge of an error will be able to correct it.
When the work goes offline and is no longer either readily correctable or protected significantly by the CDA, I agree with the concern you've expressed about liability. It's one reason I've urged that the Wikimedia Foundation stay well clear of any form of physical distribution, lest it lose all of its assets as a result. Some people wanted to make it an association of the Wikipedia authors, something problematic for similar reasons.
Online, it's pretty unlikely that it could be found liable, given the current citable precedents and the lack, so far, of ways around the clear intent and words of Congress. But it could conceivably happen that a creative lawyer finds a way.
That leads to another core structural protection: the authors are the copyright holders and owners of the works the Foundation hosts, not the Wikimedia Foundation, so the work itself and the authors are somewhat protected even if the Foundation makes a complete mess of protecting its assets. It's pretty likely that the work could continue to develop even if the Foundation vanished due to a court judgment. Not certain though.
International liability issues are frankly a mess. I doubt anyone could find a way to assure freedom from even substantial risk, probably including criminal prosecution potential for officers of the Foundaion in some jurisdictions.
It seems fairly likely, but unprovable, that few people saw it and nobody cared until Mr. Seigenthaler saw it. It was too obviously wrong and grossly incomplete to anyone familiar with his excellent body of work in the past.
You're probably right about the potential source of harm - sadly I think this really demonstrated also that few people are aware of Mr. Seigenthaler these days, unfortunate given the significance of the organisation he founded. I can hope that people are more aware of that work now, at least.
The general principle relied on at Wikipedia is that as time goes on, then umber of people who have viewed each change will increase and the chance of something inappropriate or wrong will decrease. That in turn leads to a mechanism:
1. Indicate time since last edit, as a "stability" measure.
2. Indicate total number of different authors, as an indication of breadth of consideration of the content.
Few authors and minimal time since last edit: chance of poor content relatively high.
Five thousand authors and two months since the last change: it's pretty likely that there's nothing major wrong with it. At the least, any competing viewpoints have achieved some form of mutually tolerable truce.
Seigenthaler: two authors, long time since edit: an obvious backwater. Way too small an author count to conclude that it's anything other than presumably unreliable until it's received more review. Obviously not widely examined and considered.
Not that hard to do and it avoids the need to limit who can do things, while providing an explicit indication that something is unstable and might be problematic.
1. Find reputable news source reporting the disagreement - the more reputable the better. Possibly several.
2. Change the text to use words to the effect of "The TRC concluded that whatever, a finding which is disputed [cite dispute with source].
3. Include the source link in the references and also if useful in your edit comments.
It's a lot less likely that a well sourced edit will be reverted and if it is, the next stop is the talk page to point out that all substantial views on a topic are supposed to be covered, as part of the general neutral point of view policy. Include several more references to the dispute as part of that, to make it obvious to all readers that it's not just you with a personal view.
In the event that that is unsuccessful, the next stop is using the peer review request mechanism to involve a wider part of the community in the discussion.
Which language version of Wikipedia did this happen in? The current and all of the subset of versions I checked of the English language article did not mention an ANC bombing.
A person played a silly joke at work and hurt Mr. Seigenthaler's feelings. Understandable enough reaction and he has my sympathy for the affect it had on him.
The chance of it humiliating Mr. Seigenthaler before his friends seems approxiimately zero. I assume that his friends knew that it's routine for investigations to consider all contacts with a victim and that it's ridiculous to suggest that he spent years living in Russia. Yet I can still understand him being hurt by those suggestions, given his relationship and the anti-Soviet culture in the US for much of his life.
What's not so understandable is then leveraging that to have what might be a permanent effect on what's one of the more useful information sources on the web, one which is rapidly increasing in both quantity and quality, even though it's definitely imperfect still.
That's tossing out the information resource with the prankster, like dismissing the New York Times as a whole because one reporter fabricated news stories which got past all of its quality controls. Like the Times, Wikipedia and its users are victims, not just Mr. Seigenthaler.
Obviously it's the end of the wiki anyone can edit, because anyone can no longer edit those parts of it. It's not the beginning of that because it already happened a few weeks ago, with the recent "experimental" ending of the 14,000 new page creations a month by those without an account (about 1/3 of all new pages). That's likely to have a far larger effect on decreasing content creation and improvement.
Possible negative consequences include creeping de-wikification, if this spreads to pages which are called "finished" or just spreads to a lot of pages.
Possible positive effects include reduced vandalism, though if a few pages are affected, it seems unlikely to have a significant effect on total vandalism levels.
So long as it is contained to a hundred or two pages it seems unlikely that semi-protection will do significant harm. It is likely to decrease the chance of seeing silly vandalism on a few hot target pages.
Personally, I'm more worried about one person choosing to discard 14,000 pages a month based on the story of the day. It seems fairly unlikely, unfortunately, that we'll see Mr. Seigenthaler apologising for the lasting harm he's indirectly caused by provoking that reaction over a silly joke making unbelievable claims about him. So, the correctable and somewhat quality-controlled version of the web is that much weaker.
For anyone who missed it in the fuss at the time: the offensive content in the Seigenthaler article was first removed by an anonymous contributor. What one put in, another removed. Which is exactly how it's supposed to work.
The summary you quoted said that "Pricing/booking" is done by the MySQL servers. You don't think telling people the prices and letting the people book their trips is mission-critical? It seems unlikely that they would earn much money if they didn't let anyone make bookings and get those credit card numbers from them.
What Sabre said about MySQL is "To provide the best possible service requires maintaining a massive database that can meet extreme performance demands, with very high reliability. The MySQL database has easily met these needs, and for a low total cost of ownership" and "We benchmarked our application on several databases, including open source, commercial and a specialized in-memory relational database, and MySQL was the best performing database". From http://www.mysql.com/news-and-events/press-release /release_2003_33.html
When it comest to AdWords, an "unmitigated disaster" that even "heroic performance tuning" couldn't save, even after rewriting every query for it, followed by reverting to the better-performing MySQL solution, is not what I'd call a resounding success story for a migration to a purely commercial database. Zack has some useful quotes and other links at http://zurlocker.typepad.com/theopenforce/2005/12/ googles_use_of_.html (he's MySQL's VP of Marketing).
Notice the consistency here: for both Sabre and Google, MySQL was the best performing solution, not the databases they compared it to.
One manager tried going from MySQL to commercial and had a disaster, even after rewriting everything, while the other went from commercial to MySQL and saved their company millions of dollars in TCO while delivering a world-class solution. I know which manager I'd rather be!:)
I've done the multiple database support routine myself. Best to know the databases you're going to support in advance. Standards might as well not exist when you go for maximum efficiency. Fortunately I generally didn't need maximum efficiency and the worst I generally had to deal with was Oracle limitations compared to Sybase and Microsoft SQL Server.
Even with the success stories here, there really are some cases where MySQL doesn't provide the best business case - but it's getting harder and harder to find them. There always will be some, though.
Not quite. It went down due to two circuit breakers failing, then two different brands of battery backed up disk controller not having turned off the drive write buffer, thereby losing what they had the battery to protect: data which was already written. That corrupted 5 of the database servers. We recovered from another.
Google probably considers AdWords to be mission-critical, since it is their revenue source. Ditto for Sabre/Travelocity, since the MySQL-driven travel systems are how they make their money.
Saying that MySQL isn't used for mission-critical enterprise systems isn't reflecting reality as it is today, any more than the old claim that it didn't have transactions, when these days it ships as standard with three engines supporting transactions.
I released the information that it was the hard drive controllers. Two different brands. InnoDB if it said anything was relaying what I said. Similar incident at LiveJournal, we used the same controllers as them, they did very extensive tests and demonstrated that the controllers weren't making sure that the data in the drive buffers was getting to the disk surface. Both subsequently released firmware updates.
MySQL performance is vital. It's doing in the ballpark of a billion queries per day even while more than 80% of the pages are being served by non-database cache.
Benchmarks use whatever seems best according to the requirements of that benchmark. Wikipedia uses InnoDB for everything significant and LiveJournal is one place which saw an increase in performance when switching from MyISAM to InnoDB. Depends on the workload.
I expect everyone to use the mixture which makes sense for them. Including a different database server from MySQL if that's what they need for the features they really need.
You think Google and Yahoo and Wikipedia are _small_ sites?:) There are things MySQL isn't so suitable for but when you can get to be in the top few sites in the world, that's plenty good enough for a fair chunk of the database using population.
If anyone has a MySQL database bigger than 75TB I'm interested in hearing about it - you might be the biggest in the world. So far.
I don't remember seeing a MySQL-included comparison for that 10TB dataset. Would be interesting to see how it does, particularly six months from now. Have you seen any at the larger dataset end which do include MySQL?
It's not so much that Mr. Seigenthaler didn't know how to identify the user but that the law makes it quite difficult for him to do so in this situation. He would need to demonstrate some reasonable prospect of success on a libel claim in court to get a judge to order that the ISP turn over the records of who had that IP address. The person who had that address could make a defence as a John Doe objecting to that disclosure, making it less likely to succeed.
Without such an order, privacy laws generally require the DSL provider not to disclose the details of their customer.
Wikipedia does not keep a full history of all versions. I haven't checked but if I recall correctly the versions with the text Mr. Seigenthaler posted in newspaper has been removed from the history of the article.
Wikipedia does keep a history which is almost complete, absent a request to remove part of the history for some reasonable-seeming cause.
I would appreciate it if you would convey the following to Mr. Seigenthaler, though I'll certainly understand if he's tired of the whole business by now or doesn't want to address some of the problems. I'm interesed in his views on the matters. If he's still considering legal action I'll fully understand his inability to comment on some of the matters.
His libel claim about the incident at Wikipedia is interesting for a variety of reasons:
1. Do you think that anyone who knows Mr. Seigenthaler would have believed the text in Wikipedia? He and others appear to be asserting that it's not a reputable source. At least one significant legal decision has ruled that libel from a non-credible source isn't possible, because there has to be some potential for harm and if it's not believable it can't do that harm. You know him. Would you have believed it? If not, what is the basis for a successful libel claim?
2a. Please ignore your actual intended posting for the purpose of this question. Is Mr. Seigenthaler of the Jewish religion? (a rhetorical question, I don't care about or solicit the answer) I ask because you called him a "gentile" and if he's of the Jewish religion and considers it particularly significant he might consider that to be libelous. Assuming for the moment that he considers your statement to be libelous, do you and/or Mr. Seigenthaler believe that Slashdot.com should be liable for your act?
2b. Assuming that you really meant "gentle" instead of "gentile", what regress should Mr. Seigenthaler have against you? Against Slashdot? Should asking for you or Slashdot to correct it be sufficient to correct your mistake if either party does make the correction? Slashdot is protected by the Communications Decency Act, even if it completely ignores requests for a correction, but you aren't... so what liability do you think you should have? Should Slashdot be protected from your mistake? Would it be a viable business if it was liable for your mistake?
3a. It's commonly required in many jurisdictions for the party claiming libel to first request a retraction and if such a retraction happens, there are commonly no grounds for a libel case. Is Mr. Seigenthaler in a jurisdiction where this is mandatory? Is the place where Mr. Seigenthaler suggests that Wikipedia exists or where the person claimed to have libeled him is? Would Mr. Seigenthaler be satisfied with such a retraction, which appears to have already happened on the Wikipedia.org side?
3b. Does Mr. Seigenthaler consider that the ability to remove a possible libel within seconds of seeing it is in itself sufficiently close to meeting the legal requirement for a retraction or can the party alleging libel ignore the ability to get their own correction or retraction published immediately in the same place as the claimed libel?
4a. Does Mr. Seigenthaler consider himself to be a public figure or limited-purpose public figure in respect of the matters involved in the text at wikipedia.org? If yes, does he believe that wikipedia.org or any person other than the person who wrote the text had any actual malice toward him? Given the CDA, does Mr. Seigenthaler believe that there was negligence on the part of wikipedia.org? If the answer to the last question is yes, how does Mr. Seigenthaler believe public online forums, message boards wikis and such should be treated? Should prior review of all proposed content required?
4b. At least one television channel apparently carried video of a U.S. President accidentally baring the breasts of a young woman. Does Mr. Seigenthaler believe that the television station which did so has any legal liability for doing so without its knowledge or for not having the technology in place to provide prior review of what was to be transmitted during that live event (a delay system, for example)? Do you? How could or should the TV station have anticipated the act or the possibility of, say, a libelous banner being placed in front of a camera by another participant? If the answer differs from that at
Then, some of us use MySQL for LAMP web sites doing 5,000 page views per second, not just 10,000 per day, and are approaching a billion queries per day with hundreds of gigabytes of data. Saying it's unsuitable for LAMP applications isn't very consistent with those practical results.
Did you analyze table to update the optimiser statistics after adding the rows? Did you have a suitable index? What did EXPLAIN SELECT.. say? Did you try a covering index? Did you try InnoDB, given the lack of any use of a fulltext index in that query? Were you using the program defaults (which are for tiny applications) or did you use one of the sample my.ini files?
The "smell test" is a human-understandable way of suggesting that something has been seen many times before, so why is someone filing a patent on it and is there really novelty in the application. Law aside, we probably all have some concept of what's going to fail such a test.
In this field, some improved definition of obviousness seems to be required. I would say that the fundamental purpose of the patent system is clearly expressed in the Constitution as "To promote the Progress of Science and useful Arts" and that I'm coming to the view that in this field it is doing the opposite. Not exclusively, for I do wish there was some way to reward some of the true invention of really new methods which happens.
The RIM/Blackberry case which is in the press at present is one example of a problematic patent, IMO. After reading the patent initially, I found that I'd encountered and knew of the principles being patented (yes, I'd say they were principles) in wide application for the same purpose several times, and that all of significance already existed well before the patent application was filed. It didn't come close to exceeding the standard for obviousness I'd want, because all it was doing was applying well known techniques to a device small enough to fit in a human pocket instead of a ships radio room, telegraph office or desktop email server. Ask me or others to design a system architecture for the same task and it's nearly certain that the same principles would have been used in essentially the same way.
Slashdot Mirror
Do remember that fire safety codes typically require a data center to cut all power, including all UPS power, if the EPO switch is used. That should take down any single-location site which is following those codes.
It's not the database server but the fire requlations which mandate this behavior.
Options include ignoring the law, operating in places with different law and using multiple locations.
Both places had and have a one data center vulnerability and can't really expect five nines reliability, since that's above what a single tier IV data center can be expected to assure - limit is 99.995%. Both are planning to move away from that single site vulnerability. Wikipedia is still in two to three nines territory, and in a colo which is providing tier II-III capability, and likely to stay that way for two or more years.
Both probably know what it'll take to get five nines in system architecture. Wikipedia, me included, certainly does, but isn't currently trying. It's still using donations mainly to keep up with growth on a high cost-effectiveness path, not a very high reliability path. OK enough for its route to the top few sites on the net funded by donations but you wouldn't want to do it with so little funding if downtime or less than ideal performance cost money.
The MySQL libraries are not exclusively GPL licensed, so it's not mandatory that an application using them be GPL licensed. A list of 20 other open source licenses which are accepted by MySQL for use with its client libraries can be found at http://www.mysql.com/company/legal/licensing/foss- exception.html . The PHP license is one of them.
MySQL does have two employees in Brazil, one the manager of the group which does most of the bugs verification work. Not the same as the whole company but good anyway.
It's been this way (except XA, added in 5.0) for years now. I suggest trying it yourself.
You might also find the blog aggregator at Planet MySQL interesting.
I disagree about the significan attack vector for vandalism, because as you said "these pages can remain vandalized for months until someone comes along and finds the page". If nobody is finding it, it's insignificant because nobody is being harmed by what they aren't seeing. It does take a degree of understanding of the nature of the work and its building process, and the self-restraint and judgment not to over-react when finding something obviously ridiculous.
It is good to try to improve quality assurance efforts for such things, though.
The key reason given for the "experiment" of no anonymous page creation was a lack of inward links to such articles. One alternative approach would have been to specifically target that: prevent new page creation only for unlinked articles. That would more specifically target the expressed problem and reduce the large side-effect of a total block.
A one third reduction in the rate of new article creation isn't a slight reduction. Remember that the expressed goal of the project is to contain the sum of human knowledge. A million pages is a tiny proportion of that - we're still in very early days of the project's growth and there's a tremendous amount of specialised and even quite general knowledge which isn't present yet.
So far in all the reports of his words I've seen, he's reported that the people he mentioned it to didn't believe it. Makes it pretty unlikely he was humiliated.
Agreed that libel per se as in murder is still libel, but that's not what happened here - the closest the offensive content came was suggesting an investigation, which both of us know would have applied to everyone with even passing connection to Kennedy, such was scope of the many investigations of the matter.
A wiki of this sort has no choice but to not use prior review. Its core principle is that anyone can contribute and anyone with significant knowledge of an error will be able to correct it.
When the work goes offline and is no longer either readily correctable or protected significantly by the CDA, I agree with the concern you've expressed about liability. It's one reason I've urged that the Wikimedia Foundation stay well clear of any form of physical distribution, lest it lose all of its assets as a result. Some people wanted to make it an association of the Wikipedia authors, something problematic for similar reasons.
Online, it's pretty unlikely that it could be found liable, given the current citable precedents and the lack, so far, of ways around the clear intent and words of Congress. But it could conceivably happen that a creative lawyer finds a way.
That leads to another core structural protection: the authors are the copyright holders and owners of the works the Foundation hosts, not the Wikimedia Foundation, so the work itself and the authors are somewhat protected even if the Foundation makes a complete mess of protecting its assets. It's pretty likely that the work could continue to develop even if the Foundation vanished due to a court judgment. Not certain though.
International liability issues are frankly a mess. I doubt anyone could find a way to assure freedom from even substantial risk, probably including criminal prosecution potential for officers of the Foundaion in some jurisdictions.
It seems fairly likely, but unprovable, that few people saw it and nobody cared until Mr. Seigenthaler saw it. It was too obviously wrong and grossly incomplete to anyone familiar with his excellent body of work in the past.
You're probably right about the potential source of harm - sadly I think this really demonstrated also that few people are aware of Mr. Seigenthaler these days, unfortunate given the significance of the organisation he founded. I can hope that people are more aware of that work now, at least.
The general principle relied on at Wikipedia is that as time goes on, then umber of people who have viewed each change will increase and the chance of something inappropriate or wrong will decrease. That in turn leads to a mechanism:
1. Indicate time since last edit, as a "stability" measure.
2. Indicate total number of different authors, as an indication of breadth of consideration of the content.
Few authors and minimal time since last edit: chance of poor content relatively high.
Five thousand authors and two months since the last change: it's pretty likely that there's nothing major wrong with it. At the least, any competing viewpoints have achieved some form of mutually tolerable truce.
Seigenthaler: two authors, long time since edit: an obvious backwater. Way too small an author count to conclude that it's anything other than presumably unreliable until it's received more review. Obviously not widely examined and considered.
Not that hard to do and it avoids the need to limit who can do things, while providing an explicit indication that something is unstable and might be problematic.
I suggest the following approach:
1. Find reputable news source reporting the disagreement - the more reputable the better. Possibly several.
2. Change the text to use words to the effect of "The TRC concluded that whatever, a finding which is disputed [cite dispute with source].
3. Include the source link in the references and also if useful in your edit comments.
It's a lot less likely that a well sourced edit will be reverted and if it is, the next stop is the talk page to point out that all substantial views on a topic are supposed to be covered, as part of the general neutral point of view policy. Include several more references to the dispute as part of that, to make it obvious to all readers that it's not just you with a personal view.
In the event that that is unsuccessful, the next stop is using the peer review request mechanism to involve a wider part of the community in the discussion.
Which language version of Wikipedia did this happen in? The current and all of the subset of versions I checked of the English language article did not mention an ANC bombing.
A person played a silly joke at work and hurt Mr. Seigenthaler's feelings. Understandable enough reaction and he has my sympathy for the affect it had on him.
The chance of it humiliating Mr. Seigenthaler before his friends seems approxiimately zero. I assume that his friends knew that it's routine for investigations to consider all contacts with a victim and that it's ridiculous to suggest that he spent years living in Russia. Yet I can still understand him being hurt by those suggestions, given his relationship and the anti-Soviet culture in the US for much of his life.
What's not so understandable is then leveraging that to have what might be a permanent effect on what's one of the more useful information sources on the web, one which is rapidly increasing in both quantity and quality, even though it's definitely imperfect still.
That's tossing out the information resource with the prankster, like dismissing the New York Times as a whole because one reporter fabricated news stories which got past all of its quality controls. Like the Times, Wikipedia and its users are victims, not just Mr. Seigenthaler.
Obviously it's the end of the wiki anyone can edit, because anyone can no longer edit those parts of it. It's not the beginning of that because it already happened a few weeks ago, with the recent "experimental" ending of the 14,000 new page creations a month by those without an account (about 1/3 of all new pages). That's likely to have a far larger effect on decreasing content creation and improvement.
Possible negative consequences include creeping de-wikification, if this spreads to pages which are called "finished" or just spreads to a lot of pages.
Possible positive effects include reduced vandalism, though if a few pages are affected, it seems unlikely to have a significant effect on total vandalism levels.
So long as it is contained to a hundred or two pages it seems unlikely that semi-protection will do significant harm. It is likely to decrease the chance of seeing silly vandalism on a few hot target pages.
Personally, I'm more worried about one person choosing to discard 14,000 pages a month based on the story of the day. It seems fairly unlikely, unfortunately, that we'll see Mr. Seigenthaler apologising for the lasting harm he's indirectly caused by provoking that reaction over a silly joke making unbelievable claims about him. So, the correctable and somewhat quality-controlled version of the web is that much weaker.
For anyone who missed it in the fuss at the time: the offensive content in the Seigenthaler article was first removed by an anonymous contributor. What one put in, another removed. Which is exactly how it's supposed to work.
The summary you quoted said that "Pricing/booking" is done by the MySQL servers. You don't think telling people the prices and letting the people book their trips is mission-critical? It seems unlikely that they would earn much money if they didn't let anyone make bookings and get those credit card numbers from them.
e /release_2003_33.html
/ googles_use_of_.html (he's MySQL's VP of Marketing).
:)
What Sabre said about MySQL is "To provide the best possible service requires maintaining a massive database that can meet extreme performance demands, with very high reliability. The MySQL database has easily met these needs, and for a low total cost of ownership" and "We benchmarked our application on several databases, including open source, commercial and a specialized in-memory relational database, and MySQL was the best performing database". From http://www.mysql.com/news-and-events/press-releas
When it comest to AdWords, an "unmitigated disaster" that even "heroic performance tuning" couldn't save, even after rewriting every query for it, followed by reverting to the better-performing MySQL solution, is not what I'd call a resounding success story for a migration to a purely commercial database. Zack has some useful quotes and other links at http://zurlocker.typepad.com/theopenforce/2005/12
Notice the consistency here: for both Sabre and Google, MySQL was the best performing solution, not the databases they compared it to.
One manager tried going from MySQL to commercial and had a disaster, even after rewriting everything, while the other went from commercial to MySQL and saved their company millions of dollars in TCO while delivering a world-class solution. I know which manager I'd rather be!
I've done the multiple database support routine myself. Best to know the databases you're going to support in advance. Standards might as well not exist when you go for maximum efficiency. Fortunately I generally didn't need maximum efficiency and the worst I generally had to deal with was Oracle limitations compared to Sybase and Microsoft SQL Server.
Even with the success stories here, there really are some cases where MySQL doesn't provide the best business case - but it's getting harder and harder to find them. There always will be some, though.
Not quite. It went down due to two circuit breakers failing, then two different brands of battery backed up disk controller not having turned off the drive write buffer, thereby losing what they had the battery to protect: data which was already written. That corrupted 5 of the database servers. We recovered from another.
Google probably considers AdWords to be mission-critical, since it is their revenue source. Ditto for Sabre/Travelocity, since the MySQL-driven travel systems are how they make their money.
Saying that MySQL isn't used for mission-critical enterprise systems isn't reflecting reality as it is today, any more than the old claim that it didn't have transactions, when these days it ships as standard with three engines supporting transactions.
I released the information that it was the hard drive controllers. Two different brands. InnoDB if it said anything was relaying what I said. Similar incident at LiveJournal, we used the same controllers as them, they did very extensive tests and demonstrated that the controllers weren't making sure that the data in the drive buffers was getting to the disk surface. Both subsequently released firmware updates.
MySQL performance is vital. It's doing in the ballpark of a billion queries per day even while more than 80% of the pages are being served by non-database cache.
Benchmarks use whatever seems best according to the requirements of that benchmark. Wikipedia uses InnoDB for everything significant and LiveJournal is one place which saw an increase in performance when switching from MyISAM to InnoDB. Depends on the workload.
I expect everyone to use the mixture which makes sense for them. Including a different database server from MySQL if that's what they need for the features they really need.
You think Google and Yahoo and Wikipedia are _small_ sites? :) There are things MySQL isn't so suitable for but when you can get to be in the top few sites in the world, that's plenty good enough for a fair chunk of the database using population.
If anyone has a MySQL database bigger than 75TB I'm interested in hearing about it - you might be the biggest in the world. So far.
I don't remember seeing a MySQL-included comparison for that 10TB dataset. Would be interesting to see how it does, particularly six months from now. Have you seen any at the larger dataset end which do include MySQL?
No. It might if it hit you instead of the ground.
You really want to say that Google bodged its way to success? OK...
Time for a rehash of the C vs Pascal arguments?
Your "what good is?" question has a simple answer:
Good enough to build the most well known ad network on the planet.
Also good enough, and used for, telephone and ISP billing records and one of the better known airline booking systems.
There's ample room to criticise on theoretical grounds but also no shortage of real world business succes stories.
It's not so much that Mr. Seigenthaler didn't know how to identify the user but that the law makes it quite difficult for him to do so in this situation. He would need to demonstrate some reasonable prospect of success on a libel claim in court to get a judge to order that the ISP turn over the records of who had that IP address. The person who had that address could make a defence as a John Doe objecting to that disclosure, making it less likely to succeed.
Without such an order, privacy laws generally require the DSL provider not to disclose the details of their customer.
Wikipedia does not keep a full history of all versions. I haven't checked but if I recall correctly the versions with the text Mr. Seigenthaler posted in newspaper has been removed from the history of the article.
Wikipedia does keep a history which is almost complete, absent a request to remove part of the history for some reasonable-seeming cause.
I would appreciate it if you would convey the following to Mr. Seigenthaler, though I'll certainly understand if he's tired of the whole business by now or doesn't want to address some of the problems. I'm interesed in his views on the matters. If he's still considering legal action I'll fully understand his inability to comment on some of the matters.
His libel claim about the incident at Wikipedia is interesting for a variety of reasons:
1. Do you think that anyone who knows Mr. Seigenthaler would have believed the text in Wikipedia? He and others appear to be asserting that it's not a reputable source. At least one significant legal decision has ruled that libel from a non-credible source isn't possible, because there has to be some potential for harm and if it's not believable it can't do that harm. You know him. Would you have believed it? If not, what is the basis for a successful libel claim?
2a. Please ignore your actual intended posting for the purpose of this question. Is Mr. Seigenthaler of the Jewish religion? (a rhetorical question, I don't care about or solicit the answer) I ask because you called him a "gentile" and if he's of the Jewish religion and considers it particularly significant he might consider that to be libelous. Assuming for the moment that he considers your statement to be libelous, do you and/or Mr. Seigenthaler believe that Slashdot.com should be liable for your act?
2b. Assuming that you really meant "gentle" instead of "gentile", what regress should Mr. Seigenthaler have against you? Against Slashdot? Should asking for you or Slashdot to correct it be sufficient to correct your mistake if either party does make the correction? Slashdot is protected by the Communications Decency Act, even if it completely ignores requests for a correction, but you aren't... so what liability do you think you should have? Should Slashdot be protected from your mistake? Would it be a viable business if it was liable for your mistake?
3a. It's commonly required in many jurisdictions for the party claiming libel to first request a retraction and if such a retraction happens, there are commonly no grounds for a libel case. Is Mr. Seigenthaler in a jurisdiction where this is mandatory? Is the place where Mr. Seigenthaler suggests that Wikipedia exists or where the person claimed to have libeled him is? Would Mr. Seigenthaler be satisfied with such a retraction, which appears to have already happened on the Wikipedia.org side?
3b. Does Mr. Seigenthaler consider that the ability to remove a possible libel within seconds of seeing it is in itself sufficiently close to meeting the legal requirement for a retraction or can the party alleging libel ignore the ability to get their own correction or retraction published immediately in the same place as the claimed libel?
4a. Does Mr. Seigenthaler consider himself to be a public figure or limited-purpose public figure in respect of the matters involved in the text at wikipedia.org? If yes, does he believe that wikipedia.org or any person other than the person who wrote the text had any actual malice toward him? Given the CDA, does Mr. Seigenthaler believe that there was negligence on the part of wikipedia.org? If the answer to the last question is yes, how does Mr. Seigenthaler believe public online forums, message boards wikis and such should be treated? Should prior review of all proposed content required?
4b. At least one television channel apparently carried video of a U.S. President accidentally baring the breasts of a young woman. Does Mr. Seigenthaler believe that the television station which did so has any legal liability for doing so without its knowledge or for not having the technology in place to provide prior review of what was to be transmitted during that live event (a delay system, for example)? Do you? How could or should the TV station have anticipated the act or the possibility of, say, a libelous banner being placed in front of a camera by another participant? If the answer differs from that at
Then, some of us use MySQL for LAMP web sites doing 5,000 page views per second, not just 10,000 per day, and are approaching a billion queries per day with hundreds of gigabytes of data. Saying it's unsuitable for LAMP applications isn't very consistent with those practical results.
Did you analyze table to update the optimiser statistics after adding the rows? Did you have a suitable index? What did EXPLAIN SELECT.. say? Did you try a covering index? Did you try InnoDB, given the lack of any use of a fulltext index in that query? Were you using the program defaults (which are for tiny applications) or did you use one of the sample my.ini files?
The "smell test" is a human-understandable way of suggesting that something has been seen many times before, so why is someone filing a patent on it and is there really novelty in the application. Law aside, we probably all have some concept of what's going to fail such a test.
In this field, some improved definition of obviousness seems to be required. I would say that the fundamental purpose of the patent system is clearly expressed in the Constitution as "To promote the Progress of Science and useful Arts" and that I'm coming to the view that in this field it is doing the opposite. Not exclusively, for I do wish there was some way to reward some of the true invention of really new methods which happens.
The RIM/Blackberry case which is in the press at present is one example of a problematic patent, IMO. After reading the patent initially, I found that I'd encountered and knew of the principles being patented (yes, I'd say they were principles) in wide application for the same purpose several times, and that all of significance already existed well before the patent application was filed. It didn't come close to exceeding the standard for obviousness I'd want, because all it was doing was applying well known techniques to a device small enough to fit in a human pocket instead of a ships radio room, telegraph office or desktop email server. Ask me or others to design a system architecture for the same task and it's nearly certain that the same principles would have been used in essentially the same way.