On Mon, Dec 03, 2001 at 07:25:41PM +0100, Emiliano wrote:
> > thread in cyrillic@ raised some problems with strictness of article names
> > (non-emptiness). There are several sites already in production where
> > emptiness of article names was used as indicator of article's belonging.
> > And this kind of legacy stops us from more revolutional changes as well.
> And at some point we'll have to deal with them. Rather than keeping them it'd 
> be more productive to talk about how to solve the issue behind it. And having 
> fields do double-duty (with a special state of the variable having a specific 
> meaning) is a well-known oops-shouldn't-have-done-that (like the y2k issue). 
> It's something you want to find a good solution for (and yes, that includes a 
> good solution for those currently using it that way), not something you want 
> to have around forever.
That's true. Not for point-versions though.

> Parameters are easily covered by referential integrity checks. And until 
> something better pops up, SQL is the way to go for data storage.
This does not means that referential integrity is the only way to solve
this task. For example, I would like to see our tree-like structure better
modelled using nested sets instead of adjacency model. This is covered in
old DBMS Magazine's article (circa '96) a link to which I can't find now but have
one for its new edition:

http://www.intelligententerprise.com/001020/celko1_1.shtml

this will make tree management much more easier for all components, but it 
is hard to put it into refint way for in-table relations between objects
which we have almost everywhere and which are in great interest because
orphaned branches harder to find than orphaned links to different kind of
objects.

-- 
/ Alexander Bokovoy
$ cat /proc/identity >~/.signature
  `Senior software developer and analyst for SaM-Solutions Ltd.`
---
Nov 21 20:58:58 alconost kernel: VFS: Busy inodes after unmount. 
                    Self-destruct in 5 seconds.  Have a nice day...

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to