Ports Future [was Re: Panther tickets]

[Jordan K. Hubbard]

On May 20, 2009, at 11:06 PM, Scott Haneda wrote:

> Hijacking this to a new thread.  First time posting to the dev list,  
> be gentle...
>
> On May 20, 2009, at 8:43 PM, Jordan K. Hubbard wrote:
>
>> 1. Every #ifdef or Panther-only work around adds to the overall  
>> support burden of MacPorts.  Some day, assuming MacPorts lives long  
>> enough to have such problems, most of the current set of people  
>> will be retired / MIA / dead and it will fall to a new crop of  
>> engineers to support the aging ball of goop collectively known as  
>> MacPorts.
>
> This sounds ominous.  I push ports over fink all the time.  I never  
> really knew why, until one day.  I saw there were a few apple.com  
> email addresses on this list, and I also saw it was part of  
> MacOSForge, which has a tie in to Apple in some way.
>
> I translated this as meaning there was at least a stronger chance of  
> ports lasting the test of time than anything else.

Anything is possible, though also bear in mind the fact that we  
probably have different time-scales in mind for those "ominous words"  
I uttered.  When I said "some day", I meant "sometime in the next 20  
years" since those are the time scales in which I consider software  
(unless said software is a computer language, in which case double  
that) living a full and complete life-span.

I don't see that MacPorts is in any imminent danger, no.  My  
prediction is rather that it will continue to grow and slowly morph as  
it tracks various Mac OS X releases from Apple.  It's specifically the  
fact that it has to "impedance match" for each OS release differently  
that makes the length of its legacy tail even more critical than it  
would be for most software systems, since that matching extends from  
the base infrastructure all the way out into the various ports which  
sometimes need to specify different platform blocks for each release  
variant they're expected to support.  That's a big ship to try to send  
in 4 directions at once.

If it were me calling the shots, and it's not, what I'd probably do is  
fork the entire macports tree with a -legacy branch, the purpose for  
which is defined as "everything from Panther back to Cheetah, if  
people really want", and leave -trunk advancing along with a "whatever  
the last n releases are" support policy, older releases dropping off  
into the -legacy branch to die a slower, semi-supported death.  Then,  
at least, those poor unfortunates who are constrained to $ 
{someOldRelease} of MacOSX have the option of self-support by being  
given a place to work.

- Jordan