Custom fields schema

[Myk Melez]

Christopher Hicks wrote:

> There's enough text being sent around about this without straw man 
> arguments such as this.  If none of those folks are doing custom 
> fields as extensively as Sean is then they're not relevant to this 
> discussion.

I suspect that the majority of installations that use custom fields 
don't use them nearly as extensively as Sean, but that doesn't make them 
irrelevant, especially if they are the large majority.

Nevertheless, the hundreds of Bugzilla installations I was referring to 
include all those making extensive use of standard fields, all of which 
are built on FAC, and which show FAC working well in extensive use.

> That leaves one working example that applies to the question of how to 
> implement custom fields and that is Sean's.  Given the choice between 
> betting on the horse going around the track and the horse in your 
> imagination, my money is on Sean.

Sean's implementation of custom fields is certainly not the only one, 
although it may be the most extensively developed.  But if "the 
alternative is all in my imagination" is the anti-FAC argument-du-jour, 
fine, I've attached an alternative implementation to the bug.

> Beyond that, I don't think nearly as much credit has been given to 
> Sean for (A) making something that works and being willing to share it 
> or (B) his persistence in battling with bugzilla developers who seem 
> to ignore A.

Sean deserves much credit for making something that works and being 
willing to share it, and he's demonstrated valuable persistence in 
arguing his case, but it's a gross misstatement to say that Bugzilla 
developers have been ignoring it.  The problem is what works for one 
installation at the moment is not necessarily the same as what works for 
all Bugzilla installations in the long term, and it's important to give 
those issues due consideration, especially with a feature that has so 
many far-reaching and long-term implications for the codebase.

> In this whole FAC vs. FAR debate I think I've missed something.  In 
> Sean's discussion of Myk's proposal he indicated that Myk was 
> proposing that each new custom field would be in its own table.  Is 
> that really so?  Am I the only person who finds that bizarre?  I can 
> understand FAR being a tough pill for a relational die hard to 
> swallow, but coming up with something even more outlandish as an 
> alternative is bizarre.

No, I wasn't proposing that, but I did propose putting sparse fields 
into their own tables, which is part of the reason multi-tabular 
database systems were designed in the first place.  There's nothing 
outlandish about it at all.

-myk