MacPorts caching of distfiles [was Re: ntfs-3g 1.1120 source missing]

[Rainer Müller]

Jordan K. Hubbard wrote:
>> All distfiles in one single directory? I am against that at all!
> 
> Why?  Collisions?   If so, please name the collisions in question,  
> because I cannot find any.

Maybe not yet, but maybe they will come in later? Why not be collisions 
aware?

> If you add indirection to this then you lose the ability to have a  
> global URL path that points to a mirror (either that or you need to  
> add extra logic to the MacPorts fetch code).

Why are you against adding more logic to the fetch code?

>> <group> would be category, I think. But some ports also use the same  
>> distfiles by specifying distname (e.g. vim and vim-app). So maybe it  
>> would be better to use the distname of the regarding port to avoid  
>> mirroring files multiple times.
> 
> Or you could just use a flat namespace... :-)

And have one big cluttered directory without any links which file 
belongs to which port? With the distname approach, it adapts the layout 
of /opt/local/var/macports/distfiles where distfiles are currently 
stored locally.

> I fail to understand this stubborn insistence on date stamp and group  
> information for files which are simply not meant to be user visible.   
> There is no group information you need to see on the distfile mirror  
> because you won't be managing the distfile mirror.

But maybe I want to see what files are on a mirror for some specific 
port without having to look through a list of all distfiles ever 
referenced by any port.

>> And it would be no problem to prepend URL with an advanced scheme to  
>> the master_sites list (or even add it at the end if the sourceforge  
>> mirror group is present, as said earlier). I think it would be just  
>> a bit more work to do it for ports using tagged mirrors also, but  
>> surely not impossible.
> 
> This is over-engineering at its finest.

Please explain why? It would give us the possibility to prefer some
mirror lists (e.g. sourceforge) which already got a world wide location
aware mirror network. It would not be good to throw that away and use 
the one single mirror (in case even overseas) instead.

Rainer