Request for comments: port-tags

[Michael C. Shultz]

On Monday 07 November 2005 10:18, Anton Berezin wrote:
> On Mon, Nov 07, 2005 at 09:36:20AM -0800, Michael C. Shultz wrote:
> > On Monday 07 November 2005 09:23, Anton Berezin wrote:
> > > So even for two levels, what you ask pretty much does not make a lot of
> > > sense, provided I understood your idea (and please explain in to me one
> > > more time if I did not)!
> >
> > One common gripe about the ports system is the two level directory
> > structure is too flat, and certain folk resist rather vigorously
> > adding sub categories.  The sugestion put forth was to use tags as a
> > multilevel directory structure and using the actual structure only as
> > place holders
> >
> > The idea is to decouple the actual ports directory structure, and come
> > up with something more useful without having to change the actual
> > ports structure.  If your tags are only one level deep, we aren't much
> > better off than before.
>
> No argument here, more or less.  We, however, seem to disagree about the
> actual implementation.
>
> Maybe we should not think about tags as categories and dispense with the
> idea of a `level' altogether.
>
> Tags, as opposed to multi-leveled categories, are equal to each other.
> One needs to be very careful imposing a structure where there is not
> one.
>
> So instead of thinking in terms of tags as multi-level categories on
> steroids, the idea is to be more operational:
>
> "I want a _mail_ _client_ that supports _maildir_ and _imap_".  No
> levels.
>
> The actual technical implementation becomes much more sane, too. Think
> "joins", as opposed to pre-populating huge text file with all [alright;
> not all;  some;  who gets to define which ones?] possible combinations
> of tags.
>
> So, to re-iterate, the right operational mode for this is not "go there,
> see what's in it", but "impose a restriction, see what's left".
>
> Are we on the same wave?  :-)
>
> \Anton.

OK, but if someone wants to browse the port tree to see what is there, how are 
they going to do it when the initial index is 64,000 lines long? 

-Mike