[josm-dev] Spaghetti code

[Frederik Ramm]

Hi,

Igor Shubovych wrote:
> But what about small refactoring steps. Step by step the code better.
> As I see it is histrorical problem, you didn't want to broke the plugins.

There is absolutely nothing to say against stepwise improvement! If 
someone says "I need to change this because otherwise I cannot implement 
this cool new feature" then he's unlikely to be told "please don't".

However, in the past we have had people who wanted to make lots of 
internal changes to JOSM for the perceived benefit of "easier 
maintenance" or "better future development possibilities" or so. I.e. 
there was no tangible benefit, no "if I make this change then JOSM users 
will be more productive tomorrow", but instead something like "if I make 
this change then perhaps other programmers will come along and implement 
cool features and then perhaps JOSM users will be more productive".

> With current design JOSM cannot evolve fast.

See, you're one of them already ;-) you say that the design needs to be 
changed to make it possible that JOSM evolves fast. You don't have a 
concrete promise for us, you just *assume* that JOSM will evolve better 
if the design is changed. There is no proof for that; your guess is as 
good as mine.

Evolution is a dangerous thing. It works by making random changes and 
then killing off everything where the change did not bring an advantage. 
  I'd hate to see JOSM being changed and then killed ;-) so let's take 
it slow.

What I normally say is: If someone implements something new, and in the 
process of doing so, wants to change something in JOSM's design, then by 
all means do so. For example, there is a public class member somewhere 
and you implement a new feature that depends on being notified when this 
member changed. Of course you can and will make that member private and 
implement the proper accessors. (And you will also re-compile all 
plugins afterwards and see if they still compile ;-) - but you should 
not needlessly make *all* members of that class private ("since I am 
editing this file, why not go all the way..."). Unless you have a good 
reason of course.

Bye
Frederik

-- 
Frederik Ramm  ##  eMail frederik at remote.org  ##  N49°00'09" E008°23'33"