Martin Stjernholm wrote:
>"Stephen R. van den Berg" <[email protected]> wrote:
>> In order to promote sanity and consistency in this respect, might it be
>> a good idea to check bug [5386]?

>Sure, done.

Thanks.

>As usual, I find your patch a bit odd. Is there some specific reason you
>didn't fix Context.list_var, but only one of the several places where it
>is used?

Well, this has some historic and some practical reasons:
a. Due to the long periods of time I sometimes had/have to maintain
   the patches myself I learned that choosing the minimal approach that
   scratches my itch usually is easier to maintain because I have less
   conflicts with upstream updates.
b. Over the years I have learned the internal source structure of Roxen
   bit by bit.  I've never taken the time to understand more than I absolutely
   needed to get the change in behaviour I thought was needed.
   In comparison to Roxen, Pike is a bit different.
   Since I can commit changes there myself, it's more rewarding to learn
   and toy with the internal structures and to make improvements (which,
   in part reflects on point a above, where I sometimes had to wait for
   very long times before getting feedback).
   In Pike, it's often easier to make localised optimisations/changes, since
   everything is more compartimentalised.
   In Roxen things usually are part of a larger fabric, where changes in
   one part have (unexpected) effects in a large number of distant other
   parts.
c. Sometimes the complete picture of what I'm changing eludes me, because
   I don't understand that part of Roxen that well, in that case I simply
   didn't spend enough time to see the "other obvious changes".

So, all things considered, for some things in Roxen, the minimalistic approach
worked best for me.
There are times where I take the more holistic approach, but that usually
is a result of the fact that the minimalistic approach will not work
(case in point: the database-per-rxml-session patch, I still consider my
solution superior to your attempt at a solution).

Regarding point a and c, I realise time is precious, but if I get
earlier feedback, possibly in the form of what I'm doing wrong in the
patch and asking for a rewrite or improvement, then I'd learn the
structures of Roxen faster and that will improve the quality of the
patch at hand and future patches.
-- 
Stephen.

For children with short attention spans: boomerangs that don't come back.

Reply via email to