RDBMS's power tons of worthwhile sites, even our good old slashdot which
uses it for session tracking and more. And the reasons are rock-solid. RDBMS
gives you more flexibility and have more tools ready built then nearly any
other method, and they end up scaling just fine.
They even allow you to run queries against them that are stored in a text
file, perhaps a zone file, for the speed freaks who want to use the caching,
redundant, mirrored system that is DNS. Or do a hashed set of text files.
Or... You'll notices that almost all of my posts have some aspect of this in
them. Slashdot doesn't dynamically create each page for users, it just
generates them every couple of minutes... So let's suppose the request takes
10 whole seconds to process. So what. Once every 5 minutes...
We want to change something? Easy... Someone wants to use the data who has
no clue as to what's going on? Easy... Tables, fields, simple... It is fast,
flexible, and coupled with HTTP will be no trouble for existing cddb authors
to implement, since they are already doing http requests. It is clean, no
need for these crazy (in my opinion) multi db replication situations, which
is asking for trouble. Is it as close to the metal as a custom file format?
Maybe not, but any file format that anyone comes up with can probably be
generated in a couple of minutes from the database. And let's talk about
tracking conflicting title submissions, e-mail contacts with the hash. Let's
talk about taking all this stuff and putting it on a mac for someone to use
easily there.
Had a programming job at a small company. They do real time position
acquisition, and had some really neat structs that took advantage of all
kinds of built in relations that the data contained. Outside programmers
coming in could spend hours and hours trying to figure out exactly what they
were trying to do. In fact, by going through their files with a good old dos
debug (which was especially annoying because of big/little endianess as they
had some cross-platform toolkits) I spent more time correcting their
documentation then I did coding. I vowed never to do a backend file format
that wasn't transparent to anyone who looked at it for 30 seconds. If I had
some neat or repeated processing to do, I could dump to something else...
So where does this leave us? Who knows... I suspect we are all going for
clean, simple, fast, flexible...
August
-----Original Message-----
From: Alan Cox <alan nospam at redhat.com>
To: cdindex nospam at freeamp.org <cdindex nospam at freeamp.org>
Date: Wednesday, March 10, 1999 11:00 AM
Subject: Re: My toughts on this issuet
> I support the idea of the RDBMS, I think it will be vital in the end, and
> what you give up in performace you gain in flexibilty... A fully
normalized
Forget it. Im not waiting a week for a CD lookup, and people wont host
anonymous oracle server database on their pet sun starfire to keep up with
you. Not to mention the small detail that there isnt a single well audited
secure database available