Stalled custom field development

[Joel Peshkin]

Christopher Hicks wrote:

> On Wed, 30 Mar 2005, Joel Peshkin wrote:
>
>> Actually, it is not paralyzed at all unless someone is foolish enough 
>> to try to do the whole thing in one huge patch that will never land.
>
>
> When somebody has an existing working implementation it doesn't seem 
> at all foolish to wish that it could land. 


The problem is that the existing implementation was incomplete and then 
effectively abandoned when the data storage schema debate began.  In the 
meantime, rather than proceeding with the structural portions that are 
perfectly do-able and landable independently, everyone seems to be 
waiting until they can do one monster patch to do everything.  That is a 
much more arduous way to do it. 

If you look at the history of  bug 91037, you will notice that it has 
been languishing because it has not even gotten to a reviewable patch. 

The team working on database-independence went through a similar 
evolution.  There were a few attempts at doing it all-at-once that had 
monster patches that bitrotted by the time they had uploaded and always 
were missing something.  When they realized how difficult it is to do 
that in one shot, they started breaking the problem up into a bunch of 
changes to make the SQL more portable and a change to checksetup, 
etc...  It became no less than 30 bugs, most of which could be reviewed 
and landed with relative ease.  At this point, they are just about done.

There is no reason that customfields needs to be a monster patch at 
all.  A bunch of the restructuring happens totally behind the scenes.  
Then, when you add a field by any mechanism, everything "just works."  
Once you have everything "just work" for fields in the bugs table added 
by checksetup, then you can optionally add the ability for everything to 
"just work" when the fields are somewhere other than the bugs table, and 
then adding them by other means becomes merely the addition of a UI.

There is some good news.  You will also see that it recently became 
blocked by a whole bunch of smaller bugs.  Mkanat did that and he is one 
of the guys who helped get all of the database-indepndence things 
rolling.  So, in spite of the fact that the debate over the eventual 
location of the columns is raging on, many of the prerequisites are 
being worked.  Finally.

-Joel