On July 28, 2011 04:26:14 pm Tom Sharpless wrote:
> My 2 cents.

thanks for sharing.

 
> It seems to me Hugin has been doing rather well during my time with
> it.  Chaotic, sporadic, but basically going in the right direction.

Even madness has method.  Even the most finely tuned combustion engine has 
different phases, sometimes opposite (explosion vs. compression) that when 
properly synchronized and canalized yield to a seemingly linear move forward.  
Similarly, a project goes through expansion and restriction.


> I wouldn't want to see this beautiful monster turn into a well-oiled
> machine, with a clean, rational, documented code base, protected by
> lots of shalts and shalt-nots.

That's not the intention.  We have relatively lose rules and they work well.  
Almost all rules are of the "you may" kind.  I would be the first to question 
the addition of shalts and shalt-nots, and will reluctantly accept such rules 
only if the benefits obviously outweigh the drawbacks.


> keep Hugin a workable mess, dedicated to Truth, Justice and the
> Open Source Way.

For a mess to be workable, somebody has to clean up and make place for more 
mess because when the mess becomes unmanageable it is no longer fun.  We had a 
few years of party and now we need to clean up.  The Open Source way is alive 
and kicking and tangential to the quality of the rules (liberal or 
restrictive).

 
> That said, I do think we need a moratorium on new external
> dependencies, maybe even a roll-back, otherwise there is a serous risk
> of Hugin becoming unbuildable.  Perhaps the architecture committee
> should pay more attention to the 'dependencies base' than to the code
> base.

Which specific dependency do you have in mind?  yes there have been some 
additions over the years, and especially with the GSoC projects the 
dependencies were mostly questioned and  then deliberately chosen with 
criteria such as wide availability on the supported platforms; and 
standardization.  E.g. adding more boost libraries is not an issue since we 
are already boost-dependent and boost is multi-platform and well supported.

This is an area in which a little bit more restrictive rules might help.  E.g. 
currently we have threading issues; and Hugin uses both zthreads and boost 
threads.  There may be a consolidation opportunity here where one library 
should be made redundant.  But who has the time?

 
> If I were going to seriously re-jigger Hugin, I would focus on libpano
> and its horribly inadequate PT scripts.  As a practical matter we are
> no longer supporting  interoperability with the classic PanoTools, so
> we don't need PT scripts.  I suggest documenting Hugin's persistent
> information base, defining a new, extensible project file format to
> hold it.  (preferably .ini format, not XML) and writing a new libpano
> to handle it (changing as few APIs as possible, and dropping as many
> unused old ones as possible). That would do far more, in my opinion,
> to foster future improvement than any amount of re-factoring of GUI
> code.

You are welcome to define such a new project format;  make a proposition;  
write specifications;  even write all the necessary code if you have the time 
and skill.  It's the Open Source Way.  Nobody guarantees adoption, though.

Yuv

Attachment: signature.asc
Description: This is a digitally signed message part.

Reply via email to