Re: [Ming-dev] Future Stable vs Development branches?

[Klaus R. <kl...@re...>]

Hi Justin,

I see your point but I still have some concerns:

Currently I care a lot not breaking stable interfaces. Some interfaces 
will be removed in the future. Using these interfaces will issue a 
compiler warning and people have time to adopt.

If you develop for some time only in CVS the next release (say 0.5 or 
0.6) will be a bumpy ride for developers and also for users. The 
problems you experience right now are caused by exactly the same 
release/development process you propose.
Ming 0.3 is old and (kind of) stable. Very few release were made in 
between. So it was difficult to propagate changes and give users time to 
adopt.

But currently *any* release style is better then the current one.

Thanks for caring that much about MING!
    Klaus
> Klaus Rechert wrote:
> <snip>
>   
>> IMHO doing small and frequent releases is the better way. If there are new 
>> features available people should use them instead of waiting until the whole 
>> dev-cycle is finished. Otherwise we would end up in a similar situation with 
>> a very old "stable" release and a huge pile of new code in CVS with very few 
>> active users. 
>>     
>
> Hi Klaus,
>
> Have had some time to think about this, and am now pretty sure that's 
> *not* the approach we should be using for Ming. (specifically)
>
> The reason I'm thinking this way, is that since Ming is a library, its 
> real target market is other applications.  i.e. PHP, Salasaga, etc.
>
> Applications generally need something stable they can rely on.
>
> Ming has had a capable feature set for a while now (in the 0.4.0 series 
> at least), so many apps are only going to need stuff in there.
>
> When Ming has a "stable" 0.4.0, if we keep development in a 0.5.0 
> branch, then most applications can rely on Ming to work.
>
> However, if we keep on introducing changes to the 0.4.0 stream as part 
> of our standard dev process, we're *going* to break things for the apps 
> using Ming.  Even if unintentional and accidentally.
>
> I reckon that if we want Ming to be more widely used, that's not a 
> workable way to do things.
>
> Learning this pretty much first hand with Salasaga btw, so not just 
> speaking "in theory".  When I've coded Salasaga to work with Ming 0.4.0, 
> I expect it to *keep* working with Ming 0.4.0. ;->
>
> Does this make sense?
>
> Regards and best wishes,
>
> Justin Clift
>
>   
>> 	Klaus
>>     
>
>
>