Re: [boost] [graph] SoC update

From: Beman Dawes (bdawes_at_[hidden])
Date: 2007-08-30 15:58:35

Andrew Sutton wrote:
>> To those interested,		// sorrt
> 
> Oops... accidental pasting.
> 
>> I have more to say on the topic, but I will split it into a second
>> email since this is pretty long.
> 
> Since my SoC package constitutes a significant addition to the  
> Boost.Graph library, Jeremy and I felt that it would benefit from the  
> formal review process - especially the newer code that is concerned  
> with the library interface. However, we're unsure if the review needs  
> the same formality as a completely new library since this is, as  
> mentioned, a (albeit rather large) change to an existing library.
> 
> An alternate idea might be to take these changes and any new  
> Boost.Graph algorithms in the vault as a single introduction. This  
> could include:
> 
> - the planar graph suite
> - cycle ratio code
> - floyd-warshall (new params)
> 
> A third option - and actually not a terrible idea - would be to  
> branch the Boost.Graph trunk, perform the integration and then do  
> some serious housekeeping like cleaning up tests, examples,  
> documentations, making sure the interface is clean and consistent,  
> and - god forbid - putting Boost.Graph into boost::graph.
> 
> The more I think about it, the more I'm becoming a fan of the third  
> option. Unfortunately, it's also a fairly large chunk of work and  
> would take a serious commitment from the developer(s) working on it.
> 
> Any thoughts?

Both having a review process and doing housekeeping on a branch seem 
worthwhile.

I personally don't think the review has to be as formal as for a new 
library, but should follow the same basic procedure; it should be 
announced ahead of time, all the material should be available at an 
accessible location, it should run for 10 days or so, etc. Rather than 
having a review manager as such, perhaps Jeremy and Doug could moderate.

As far as a branch goes, we would like to keep the trunk more stable 
than in the past. So for a major change that might involve some 
instability for awhile, a branch sounds like a good idea until the 
changes stablize.

--Beman