Slashdot Mirror
PostgreSQL 8.1 Available
atani writes "PostgreSQL 8.1 has been posted, though not officially announced as of this moment. This release includes two-phased commits, improved SMP and overall performance, a new role system replaces the older user/group, autovacuum is now within the backend rather than a separate contrib module, and various improvements, performance enhancements, and bugfixes. " You can also read the developer notes for the popular database. One thing is clear- with the newest Postresql and MySql, you have much to choose from.
Just installed the Windows version. Just a quick couple notes here...
:)
1. Installer was seemingly faster than 8.0 version.
2. Installer could do a little bit better job of hiding all the 'options' (ISBN, Fuzzy search, etc) you can install. Put them behind an 'advanced' button or something - it's a little intimidating to see so many options at first. Also the PL language choices are odd - 'pl/Perl' and 'pl/Perl (untrusted)' ??? These are things that could probably be hidden from the majority of people just testing it out for the first time - either install everything by default, or nothing, but put some of these things behind 'advanced' tabs.
3. pgAdmin III bundled tool is 1.4 - I think I was using 1.3 last time I installed. Visually it looks a bit nicer - I'm assuming they've fixed some bugs or something similar to warrant a number change.
I'll probably get flamed for #2, but I'm just putting out some suggestions. The fact that there *is* a Windows installer at all is a good thing - I'd just like to see it improve to help reach a wider audience for future releases.
creation science book
I've been running my Blastwave packages[1] of the betas and RCs since the first snapshots became available, and I've been massively impressed with this release. Moving autovacuum into the main package is a really nice touch - all you have to do now is uncomment a few lines in postgresql.conf and it handles it all for you.
:)
The new roles system is also amazingly useful. You can set up a range of roles with a variety of permissions, and then let users "assume" those roles. So you can log in with a day-to-day account, and when you need to do some admin work just SET ROLE [name of your super-user role] and then revert back once you're done. Great if you want to give a junior DBA the ability to create databases, but not the ability to modify other things (such as creating new roles).
Congratulations to the PostgreSQL team anyway - for doing things "the right way"
[1]=http://www.blastwave.org/testing/
it seems like a simple check, in the same way when it checks for permissions, when doing a list of databases. Seems to make sense.
In a shared database server, it can be important. Although it in a way is security-through-obscurity, many would rather not have their database name 'companyfinances' visible to those with no access. Additionally, on a shared database services, you don't want your customers to know if there are 20 or 200 databases on that server (the number means nothing depending on the size anyway, but looks bad).
-M
when you see the word 'Linux', drink!
Okay, I'm going to bite: what on earth does a red wheelbarrow have to do with databases?
Save Maine's economy: write stuff down. All comments are exclusively my own, not my employer.
Full text searching has existed for a number of years and is actively used by a few Russian search engines on large volumes of content (millions of text documents), among other places.
Take a peak in the contrib directory of the source tree for tsearch.
Documentation for TSearch
Rod Taylor
PostgreSQL is rock solid, and very extensible (user-defined aggregates, user-defined procedural languages, user defined functions, triggers, user defined types, table functions, and much more). It probably also performs better in many situations due, in part, to MS SQLs locking vs. PostgreSQL's MVCC.
However, there are more tools available for MS SQL, and there is some form of multi-master replication and probably better table partitioning. MS SQL is not really a bad database, but I think PostgreSQL has it beat except on those two points.
Any real performance analysis is heavily application dependent, however. If that's what you care about, you need to do your own tests.
Social scientists are inspired by theories; scientists are humbled by facts.
Here's what I think he'll answer to your first question:
"
MySQL works fine on Slashdot. It has all the features and performance we need, it has been running flawlessly for years and we're already familiar with it, so why should we change to anything else? What makes you think there's a bullet that needs biting? Granted, PG looks neat and all, but why exchange a dollar for four quarters?
"
Here's what I think he'll answer to your second question:
"
MySQL 5 doesn't offer us any features we absolutely need (otherwise we'd be using PG, right?) We will upgrade eventually, but we have bigger fish to fry right now, and upgrading our database is not very imperative.
"
Something else he might say:
"
Running slashdot is not as simple as running a basement website that gets 3 hits per hour. Thought needs to be put into these decisions. We can't just run off and install something the day it's released.
"
1) It's free. SQL server costs 5000 per processor or 16,000 per processor depending on the version
2) It runs on every platform. SQL server only runs on windows. It's the only database in wisespread use that locks you to one operating system.
3) It has no limits on how much memory it uses. SQL server standard edition limits itself to 2 gigs as of SQL server 2K (don't know if they fixed that by now).
4) it supports text fields that are only limited by your OS and uses them extremely efficiently. These are not like SQL server blob fields but they are like HUGE text fields that can be indexed or used in aggregate functions.
5) It has user definable data types, user definable operators, user definable functions.
6) It can use perl, python, java, tcl or PG/Pqsl as it's stored procedure languages. YOu can also use C and even compile your C stored procs in with the server if you want super speed.
7) It has multi version concurrency control. This means readers never block writers, ever.
8) No lock escalation. SQL server users know the value of this, everybody else takes it for granted.
9) Lots of built in datatypes like arrays, IP address, geometric types, GIS types etc. Yes it's possible to write a query that asks "select all rectangles that contain this point" or "select all ip addresses in this address mask"
10) Support for hierarchies (in the contrib) so you can natually and intuitively model graphs without writing code or using complex self joins and such. Look up ltree.
11) A fantastic rule system. You can make anything look like a updateable recordset if you are willing to code it.
12) PostGIS.
I am just scratching the surface. I am sure I have missed some other features but that should whet you appetite.
evil is as evil does