Re: Ping: dawn, some mvl questions

Gene Wirchenko wrote:
> On 22 May 2006 14:09:04 -0700, "dawn" <dawnwolthuis_at_gmail.com> wrote:
>
> [snip]
>
> >ordered it in the first place, getting that order back. There really
> >might be some subtle, but real, productivity gains by users who enter
> >data and have their data retain their original ordering if they ask for
> >it again without specifying a new order. They can still do the sort by
>
> Or there may be some subtle but real inefficiencies because the
> DBMS has to preserve ordering when it is not needed.

Agreed. It would be great if there were some performance metrics that crossed data models so as to test such things.

> Why not do analysis and find out instead remaining clueless about
> how the data is to be used?

Definitely. However, don't ditch the possible usefulness of the ordering quickly just because your tools don't handle ordering well. Would you agree that at times there is a failure to even ask the question about the relevance of ordering when designing for an RDBMS? As mAsterdam and others have pointed out, the same failure happens in reverse when nested lists are handled better than nested sets.

> >alpha or whatever other ordering, but no information, whether
> >understood to be relevant or not up front, was lost if the data were
> >modeled as a list.
>
> So what? If the order matters, that came up in your analysis.

Sometimes the user is talked out of caring about the order when it isn't exceedingly important due to the cost of implementing a list by way of sets using the RM.

> If the order does not matter, that came up in your analysis. You did
> do analysis, right?

You betcha. Cheers! --dawn

> Sincerely,
>
> Gene Wirchenko
Received on Tue May 23 2006 - 03:00:38 CEST