mysql vs. postgresql

[Trent Shipley]

On Saturday 2003-06-14 22:26, Derek Neighbors wrote:
> On Sat, 2003-06-14 at 22:05, Trent Shipley wrote:
> > Another item to take into consideration is market share.
> >
> > Most of the time it is wiser to use the most popular application and not
> > the best.



> If this is the criteria.  Please pickup Windows 2000 and MSSQL Server.
> It is FAR more widely used.  

They are *VERY* good products.  Furthermore, as a user of both MS products and 
Linux-and-friends the MS products have the advantage of user-interfaces that 
are degrees of magnitude better than the freeware versions.  

The main problem of MS in comparison to gratis-ware is the initial price 
point.  

If you can afford them, the big commercial databases (in my book these are 
Oracle, DB2 and MS SQL Server) are a class above the freeware databases.


Enterprise level databases (Terrabytes, $20K and up.):

Top: 
Oracle, DB2.  Superior flexibility.  (Note: SAP claims this is a *draw back*.  
These products are *too* complex.)

Second tier (Only because of limited features): 
MS SQL Server.  (Transact*SQL is a glorified scripting language.  Limited or 
non-existent OO attributes.  NB! It looks like SAP-DB with a GPL licence may 
be technically competitive.  However, limited market penetration may mean 
that the cost to run SAP-DB could exceed those of SQL Server.  Not 
surprisingly SAP's attitude toward its database is KISS.  A database is where 
you put data used by middle-ware.  A DB should store and retrieve 
data--that's it.  Any complexity goes in business logic implemented by a 
middle-ware product ... like SAP.)


Mid-level databases (100's of Gigabytes, max: Free, MySQL is dual licensed. 
I'm not certain how much the non-GPL EULA costs.): 

-- MySQL is fast, cheap, and simple to a fault.  It is widely used.  It is 
well supported and documented.  It seems to be gaining market share.  Rumors 
of a deal with SAP for SAP-DB technology may result in a partial challenge to 
MS SQL Server.  Nevertheless, expect SQL Server to be a better option for 
most customers because MS can throw money at ease-of-use.  

(Remember my fellow gear-heads, for 99% of our fellow travelers software is a 
means to an end.  Hard to use software is nearly equivalent to useless 
software.  Yes, you can hire an expert, but no one likes doing that.  [I hate 
taking my car to the shop, for example.])


-- Postgresql is reliable, cheap, and has an obscenely complex feature set.  
(I have a personal love affair with this software.)  For better or worse, 
Postgres' heritage as a platform for academic database research is an organic 
part of the product.  A resulting major annoyance is that you have limited 
tuning knobs.  (Tuning a database is not *academically* interesting.)  All 
the objects for a named 'database' have to live on a single unix path-node.  
Forget about putting indexes and tables on different disks or 
disk-controlers.   Historically, security for Postgres has also been somewhat 
lax.  Again, not a concern for grad-school developers.

It is unlikely Postgres lacks something you need.  More likely what you need 
probably lacks Postgres.  Limited market share often means some critical 
(comercial) killer app has no interface to Postgress.  (In fact, many killer 
apps only interface with the major databases.  It is by no means uncommon to 
find that some business critical application has been built to work ONLY with 
MS SQL Server.)



From a biz perspective MySQL is to be prefered to Postgresql.  However, at 
this point MySQL may still lack critical features (not least being 
extesibility).   IF MySQL passes feasibility analysis use it instead of 
Postgres (for reasons of economics and business, *not* engineering.)


Note that MS has no $5000 database in the mid-range market.  This means Access 
users get heart-attack sized sticker shock when they realize they've outgrown 
their MS-Office based tools.


Tiny databases [AKA toys] (A few gigabytes max.  Ideally, no more than one 
user at a time.  Hundreds of dollars.): 

Access: Not bad given what it is designed to do.  Tends to corupt its 
datafiles, but if your SysAdmin is good about backups resulting losses are 
usually well inside the acceptable range.  The natural upgrade path is to 
MS-SQL, resulting in sticker-shock and learning-curve hell.  A product 
mid-way between Access and MS-SQL is desperately needed.  Due to obscene 
levels of market penetration Access databases can be migrated to nearly every 
product on the market.

Paradox:  Sadly, nearly dead.  Someone talk IBM or SUN into buying this 
Borland product and releasing it on Linux, preferably under GPL.

The freeware world has no real file-by-file personal database product.  This 
is often a major objection to OpenOffice.org.  True, a intermediate level 
guru can install Postgress on a laptop, but the whole point of the 
Access/Paradox type product is to minimize the need for expert level 
knowledge.