So you want the column to be NOT NULL, and yet you can't be bothered to tell the db your desired default value for that column. What EXACTLY do you expect the database to do ?
Um, I want it to create the table with a NOT NULL constraint and no default value. That's perfectly legitimate SQL.
Now a "proper" db might just moan at table creation time that you're trying to do something silly, whereas MySQL assumes you are silly and inserts it's own suggested default.
No, a proper db will create the table as instructed. There's nothing silly about creating a table that you don't want NULL values in, and saying that there's no sane default, so the user must specify a value on INSERT.
Horses for courses I would have said. Just for interest, as I'v enever tried Postgre, what does IT do ? Allow you to continue, then moan when you try and do an INSERT, or does it really stop you at creation time with a warning / error.
The former (reject INSERTS that don't specify a value for that column), which is what any decent database like Oracle, DB2, and hell probably even Microsoft SQL Server will do.
There's very little Oracle or anyone else can offer me that I don't already have with PG.
Basically, built-in multimaster replication (if you pay for the license, of course), and load-balanced clustering. If you don't need either of those, there's really no reason to pick Oracle over PG other than political.
PG is an awesome DB, and it just keeps getting better.
When was that? PostgreSQL has been using MVCC for as long as I can remember (as least since 6.0+, probably earlier), which is the same type of concurrency control used by Oracle. The implementation is a little different, but the effect is the same. Much more efficient than the locking method used by MySQL and MSSQL until fairly recently (SQL Server 2005 and InnoDB use MVCC).
although there's a BSDL version in the FreeBSD kernel that could probably be ported.
The ext2 driver in FreeBSD isn't very actively maintained, tends to lag behind the rest of the kernel, and has been the cause of various problems like panics and VFS lockups in the past. It's not recommended for serious read-write use.
Really the only platform to have good ext2 support is Linux. Honestly a least-common-denominator UFS variant would probably be usable by more systems due to Mac OS X.
This has nothing to do with PGO, and I really wish people would stop pointing fingers at it when the performance difference is brought up on the mozilla lists.
This has been going on for far longer than PGO has been enabled on the Windows builds. PGO may have increased the gap a little, but not that much.
There was a time around when GTK2 was first released that Seamonkey could be compiled against either. The difference was especially obvious then. I'm sure GTK2 has been improved since then, but compare it to Qt-based browsers on the same hardware and see for yourself. Most of the perceived slowness comes from interacting with the UI components, not rendering speed.
Clearly, the Mozilla developers just forgot to call 'gtk_widget_set_double_buffered(false);'. That's what's been gumming up the works. Much appreciated, and thanks for your informed opinion on the matter!
Sarcasm aside, that's just a single example of the myriad of inefficiencies in GTK. Especially in light of compositing window managers.
Yes, more likely it's GTK that's using X inefficiently. Especially since almost any program that uses GTK feels "slow". The more complex the widget usage, the worse it is. Compare to say, Qt apps, on the same hardware.
Before someone says it's local, no, this isn't a configuration issue. I've noticed this across many different platforms (FreeBSD, various linux distributions, HP-UX, etc etc.) on a variety of hardware.
It is a shame they did not do Firefox on Linux, Firefox on windows XP and Firefox on windows Vista, all on the same hardware. It would have been interesting to see how the underlying OS affects the performance of the browser. Then further compare IE on Vista vs Firefox on Ubuntu.
While hard numbers would be useful, it's painfully clear to anyone who's used it on both platforms that Firefox on windows is far faster than Firefox on linux. Try opening a bunch of tabs and see how sluggish it is on linux to switch between them or close one.
Personally, I blame GTK2's obsession with double buffering everything. I recall GTK1-based seamonkey builds being quite a bit faster than Firefox when they first switched FF over to GTK2. Of course you'd be mad to even install GTK1 these days, but the performance issues really need to be addressed. If I could get Konqueror without all the KDE baggage I would, for the brief time I used KDE it was always snappy and responsive.
I haven't tried plugging in a memory card formatted with ext2fs but does Windows prompt for a driver when it finds an unknown FS or simply ignore it?
No, it shows up as an unknown file system, and if you double click on the drive letter that it gets assigned, it "helpfully" offers to format it for you.
I've often thought it would be an interesting research project to modify an OS so that each application launches with its own security context that is a subset of the user's context. Sort of a derived userid that only has access to its own files (read access to program and write access to data).
The hardest part is not making it too painful for the user when they need to share data between applications. Ideally you'd have to explicitly give permission for this to happen, but it could get tiresome fast.
While SQLite3 doesn't support true concurrency, it's generally good enough for things like MythTV. The locking is only slightly more clunky than MyISAM's.
To be honest, and I'm a total pgsql fan, you usually wind up needing to edit pg_hba.conf and reload to allow external access. But yeah, it is about as easy to me as mysql.
Although, the primary deployment for MySQL seems to be running on the same machine as the web server. In a direct comparison of that case, you don't even need to change pg_hba.conf.
The default config isn't secure if you have untrusted users on the machine (i.e. hosting service), unless you used "initdb -A md5" or something similar, but it's easy to change.
My favorite auth method for server-in-a-box is "ident sameuser" over UNIX sockets. In that case it seamlessly integrates with the OS's underlying security model, so all I have to do is "CREATE USER webapp;", and the user 'webapp' can connect as itself (and only as itself) with no password.
The same thing in MySQL would have taken me thirty seconds now, and no more than 15 minutes when I was starting out. With Postgres, it took me upwards of 20 minutes when it should have taken much less time.
That's because you know MySQL, so of course something that works differently is going to be more work for you to figure out.
I've used PostgreSQL for years, when I had to set up a MySQL database for some php app it took much more than 15 minutes to figure it out and get it running. The primary problem was MySQL's obtuse user management system.
With PostgreSQL I know that it's secure by default -- the default user has no password, so even if you enable password authentication it won't work (because it has no password!). You log in locally with trusted authentication, and issue the very logical CREATE USER. Edit the self-documented config file to allow remote hosts to access the database using your preferred authenticaion method, and you're done.
With MySQL, new users are automagically created by the GRANT command?! Huh? On top of that, passwords are apparently specific to a certain host string. Bizarre. Do I need to use localhost for the actual machine name for local users? What about remote machine without a reverse DNS entry? What's the order of precedence for '%' vs a more specific name?
Oh, the default 'root' account has no password...and allows access over the network. Wonderful. Okay, so to change that do I use root@% or root@computer? How do I know I changed the right one and there isn't still some root@something entry? SHOW TABLES is easy enough, how about SHOW USERS? Nope, that's not it.
Time to check the startup guides. Well, one just has a single password change, another has 3 or 4 lines of 'delete from user...'. The reference for GRANT just has a bunch of caveats and warnings, and the "User Account Management" section goes on and on and somehow doesn't manage to tell me what I want to know.
To this day I'm not 100% sure if the MySQL install is secure. I decided my time would be better spent eliminating the MySQL-isms from the app in question so that it can run on Postgres like everything else on the server. There are some very strange queries in there - a lot of GROUP BY expressions that make no sense and aren't valid SQL. Some of it I'm not sure how it ever worked.
Bit-robbing is a bad example as it steals the LSB from every channel at a constant interval. It doesn't pick and choose which ones to reduce the bit-rate on.
That info is a little out of date, too. At least to the endpoint, PRI is the preferred method of attaching to the PSTN. You do give up one of the channels to be dedicated to signaling, but get vastly superior channel allocation.
I'm not sure what the telcos are using for the big trunks, but I'd imagine (hope) they have something a little more advanced than basic T1 by now.
Yeah, we have some HP-UX servers at work that use VxFS and it's pretty sweet. Actually HP licenses it and uses it as the base filesystem for HP-UX. You do have to buy a separate license to use the online resizing features and such. For an enterprise server, it's well worth it.
For my home stuff I use ZFS (on freebsd), which comes pretty close. About the only thing it doesn't do is shrinking of filesystems. I like the copy-on-write design and snapshot/clone functionality.
This used to work (and was the preferred solution), until Flash 9. Flash 9 causes the browser to crash at random, often before you even finish the first thing you try to watch.
On top of that Flash 9 only supports ALSA out of the box (which isn't emulated by the Linux compat layer). There's a buggy OSS addon module for flash that never seemed to work right.
Even in the "good old days" of flash 7 under linux compat, the sound would still drift out of sync if you didn't pause and resume every 30 seconds or so. On the bright side, running flash 9 under wine doesn't seem to suffer from this problem.
Slashdot Mirror
So you have never installed Oracle or Cache or DB2?
How do you configure these databases without their web interfaces?
vi and sqlplus
Same way you do when you disable enterprise manager because java is a memory hog.
He who lives by the sword ...
...gets shot by those who don't!
So you want the column to be NOT NULL, and yet you can't be bothered to tell the db your desired default value for that column. What EXACTLY do you expect the database to do ?
Um, I want it to create the table with a NOT NULL constraint and no default value. That's perfectly legitimate SQL.
Now a "proper" db might just moan at table creation time that you're trying to do something silly, whereas MySQL assumes you are silly and inserts it's own suggested default.
No, a proper db will create the table as instructed. There's nothing silly about creating a table that you don't want NULL values in, and saying that there's no sane default, so the user must specify a value on INSERT.
Horses for courses I would have said. Just for interest, as I'v enever tried Postgre, what does IT do ? Allow you to continue, then moan when you try and do an INSERT, or does it really stop you at creation time with a warning / error.
The former (reject INSERTS that don't specify a value for that column), which is what any decent database like Oracle, DB2, and hell probably even Microsoft SQL Server will do.
There's very little Oracle or anyone else can offer me that I don't already have with PG.
Basically, built-in multimaster replication (if you pay for the license, of course), and load-balanced clustering. If you don't need either of those, there's really no reason to pick Oracle over PG other than political.
PG is an awesome DB, and it just keeps getting better.
When was that? PostgreSQL has been using MVCC for as long as I can remember (as least since 6.0+, probably earlier), which is the same type of concurrency control used by Oracle. The implementation is a little different, but the effect is the same. Much more efficient than the locking method used by MySQL and MSSQL until fairly recently (SQL Server 2005 and InnoDB use MVCC).
although there's a BSDL version in the FreeBSD kernel that could probably be ported.
The ext2 driver in FreeBSD isn't very actively maintained, tends to lag behind the rest of the kernel, and has been the cause of various problems like panics and VFS lockups in the past. It's not recommended for serious read-write use.
Really the only platform to have good ext2 support is Linux. Honestly a least-common-denominator UFS variant would probably be usable by more systems due to Mac OS X.
This has nothing to do with PGO, and I really wish people would stop pointing fingers at it when the performance difference is brought up on the mozilla lists.
This has been going on for far longer than PGO has been enabled on the Windows builds. PGO may have increased the gap a little, but not that much.
There was a time around when GTK2 was first released that Seamonkey could be compiled against either. The difference was especially obvious then. I'm sure GTK2 has been improved since then, but compare it to Qt-based browsers on the same hardware and see for yourself. Most of the perceived slowness comes from interacting with the UI components, not rendering speed.
Clearly, the Mozilla developers just forgot to call 'gtk_widget_set_double_buffered(false);'. That's what's been gumming up the works. Much appreciated, and thanks for your informed opinion on the matter!
Sarcasm aside, that's just a single example of the myriad of inefficiencies in GTK. Especially in light of compositing window managers.
Yes, more likely it's GTK that's using X inefficiently. Especially since almost any program that uses GTK feels "slow". The more complex the widget usage, the worse it is. Compare to say, Qt apps, on the same hardware.
Before someone says it's local, no, this isn't a configuration issue. I've noticed this across many different platforms (FreeBSD, various linux distributions, HP-UX, etc etc.) on a variety of hardware.
It is a shame they did not do Firefox on Linux, Firefox on windows XP and Firefox on windows Vista, all on the same hardware. It would have been interesting to see how the underlying OS affects the performance of the browser. Then further compare IE on Vista vs Firefox on Ubuntu.
While hard numbers would be useful, it's painfully clear to anyone who's used it on both platforms that Firefox on windows is far faster than Firefox on linux. Try opening a bunch of tabs and see how sluggish it is on linux to switch between them or close one.
Personally, I blame GTK2's obsession with double buffering everything. I recall GTK1-based seamonkey builds being quite a bit faster than Firefox when they first switched FF over to GTK2. Of course you'd be mad to even install GTK1 these days, but the performance issues really need to be addressed. If I could get Konqueror without all the KDE baggage I would, for the brief time I used KDE it was always snappy and responsive.
Yes, but the question was: Does it prompt for a driver if you don't have one installed already?
In my experience, no, it doesn't.
I haven't tried plugging in a memory card formatted with ext2fs but does Windows prompt for a driver when it finds an unknown FS or simply ignore it?
No, it shows up as an unknown file system, and if you double click on the drive letter that it gets assigned, it "helpfully" offers to format it for you.
I've often thought it would be an interesting research project to modify an OS so that each application launches with its own security context that is a subset of the user's context. Sort of a derived userid that only has access to its own files (read access to program and write access to data).
The hardest part is not making it too painful for the user when they need to share data between applications. Ideally you'd have to explicitly give permission for this to happen, but it could get tiresome fast.
Both parties talk about change but the reality is that they keep the status quo once they get it
The status is not quo.
The world is a mess, and I just need to... rule it.
Yeah, but it changes them to DOS format when you save, with no option to keep the UNIX line endings :(
Good thing vim has a windows version.
While SQLite3 doesn't support true concurrency, it's generally good enough for things like MythTV. The locking is only slightly more clunky than MyISAM's.
-Does it silently accept CONSTRAINTS other than FOREIGN KEYs and then not use them? ...
Why is ignoring FOREIGN KEYs okay? A constraint is a constraint, IMO.
Last I checked FKs worked, if you're using InnoDB, and then only kinda (no deferred checking, which is the SQL standard default).
To be honest, and I'm a total pgsql fan, you usually wind up needing to edit pg_hba.conf and reload to allow external access. But yeah, it is about as easy to me as mysql.
Although, the primary deployment for MySQL seems to be running on the same machine as the web server. In a direct comparison of that case, you don't even need to change pg_hba.conf.
The default config isn't secure if you have untrusted users on the machine (i.e. hosting service), unless you used "initdb -A md5" or something similar, but it's easy to change.
My favorite auth method for server-in-a-box is "ident sameuser" over UNIX sockets. In that case it seamlessly integrates with the OS's underlying security model, so all I have to do is "CREATE USER webapp;", and the user 'webapp' can connect as itself (and only as itself) with no password.
I've used MySQL for years...
The same thing in MySQL would have taken me thirty seconds now, and no more than 15 minutes when I was starting out. With Postgres, it took me upwards of 20 minutes when it should have taken much less time.
That's because you know MySQL, so of course something that works differently is going to be more work for you to figure out.
I've used PostgreSQL for years, when I had to set up a MySQL database for some php app it took much more than 15 minutes to figure it out and get it running. The primary problem was MySQL's obtuse user management system.
With PostgreSQL I know that it's secure by default -- the default user has no password, so even if you enable password authentication it won't work (because it has no password!). You log in locally with trusted authentication, and issue the very logical CREATE USER. Edit the self-documented config file to allow remote hosts to access the database using your preferred authenticaion method, and you're done.
With MySQL, new users are automagically created by the GRANT command?! Huh? On top of that, passwords are apparently specific to a certain host string. Bizarre. Do I need to use localhost for the actual machine name for local users? What about remote machine without a reverse DNS entry? What's the order of precedence for '%' vs a more specific name?
Oh, the default 'root' account has no password ...and allows access over the network. Wonderful. Okay, so to change that do I use root@% or root@computer? How do I know I changed the right one and there isn't still some root@something entry? SHOW TABLES is easy enough, how about SHOW USERS? Nope, that's not it.
Time to check the startup guides. Well, one just has a single password change, another has 3 or 4 lines of 'delete from user...'. The reference for GRANT just has a bunch of caveats and warnings, and the "User Account Management" section goes on and on and somehow doesn't manage to tell me what I want to know.
To this day I'm not 100% sure if the MySQL install is secure. I decided my time would be better spent eliminating the MySQL-isms from the app in question so that it can run on Postgres like everything else on the server. There are some very strange queries in there - a lot of GROUP BY expressions that make no sense and aren't valid SQL. Some of it I'm not sure how it ever worked.
Bit-robbing is a bad example as it steals the LSB from every channel at a constant interval. It doesn't pick and choose which ones to reduce the bit-rate on.
That info is a little out of date, too. At least to the endpoint, PRI is the preferred method of attaching to the PSTN. You do give up one of the channels to be dedicated to signaling, but get vastly superior channel allocation.
I'm not sure what the telcos are using for the big trunks, but I'd imagine (hope) they have something a little more advanced than basic T1 by now.
Yes, the separate license is sold as "Online JFS" (JFS = VxFS, HP likes to arbitrarily rename stuff for no good reason)
Yeah, we have some HP-UX servers at work that use VxFS and it's pretty sweet. Actually HP licenses it and uses it as the base filesystem for HP-UX. You do have to buy a separate license to use the online resizing features and such. For an enterprise server, it's well worth it.
For my home stuff I use ZFS (on freebsd), which comes pretty close. About the only thing it doesn't do is shrinking of filesystems. I like the copy-on-write design and snapshot/clone functionality.
BugMeNot: it's dot com!
This used to work (and was the preferred solution), until Flash 9. Flash 9 causes the browser to crash at random, often before you even finish the first thing you try to watch.
On top of that Flash 9 only supports ALSA out of the box (which isn't emulated by the Linux compat layer). There's a buggy OSS addon module for flash that never seemed to work right.
Even in the "good old days" of flash 7 under linux compat, the sound would still drift out of sync if you didn't pause and resume every 30 seconds or so. On the bright side, running flash 9 under wine doesn't seem to suffer from this problem.