MOVED - from == to?

[Michael C. Shultz]

On Sunday 15 January 2006 17:55, Kris Kennaway wrote:
> On Sun, Jan 15, 2006 at 05:07:35PM -0600, Mark Linimon wrote:
> > On Sun, Jan 15, 2006 at 12:48:28PM -0500, Dan Langille wrote:
> > > Can the FROM field be the same as the TO field in ports/MOVED?  I
> > > think not.
> >
> > eik and I were trying to convince people that that would be a good
> > technique to note 'previously removed port has been reinstated', but I
> > don't think we convinced anyone.
> >
> > My original interpretation of MOVED, when I built portsmon, was that
> > it contained a complete historical record of ports that had been
> > deleted and/or moved.  To build the view of the 'latest' change, portsmon
> > walks the tree and follows multiple entries.  The reason I did it this
> > way is that so I could say 'PR #xyz was orginally about foo/bar-devel but
> > now it is about foo/bar since that's where foo/bar-devel got moved to."
> >
> > However, I also failed to convince people that keeping the historical
> > entries was useful, and now some (but not all) of them have been flushed.
> >
> > Given that, we might as well flush the rest of them, and the from=to
> > entries as well.
>
> There's a larger problem which is that it's impossible to correctly
> parse the information in MOVED as it currently stands.  e.g.
>
> editors/emacs|editors/emacs19|2004-03-20|emacs 19.x moved to a non-default
> port location editors/emacs21|editors/emacs|2004-03-20|emacs 21.x moved to
> default port location
>
> The intention of the above is that old versions of editors/emacs
> (which were emacs 19.x) should be followed to editors/emacs19.  But a
> newer editors/emacs port is 21.x, so it should not follow there.
>
> The way MOVED was intended to be parsed was to start at the
> "appropriate date" in the file and only parse entries after this date.
> The problem is that there's no unambigious way to determine
> "appropriate date": it's supposed to mean the date of the port from
> which the package was built, but this is not recorded in the package
> (or clearly defined itself: in most cases it's the date of the port
> Makefile, except for ports that .include other things,
> i.e. potentially every port).
>
> A better solution might be to additionally record the PKGVERSION of
> the port at the date of the MOVED entry, since this may be easily
> parsed and used to determine whether the entry applies to a given
> package.  The problem is that PKGVERSION is not always uniquely
> defined and may sometimes depend on OSVERSION and other factors.
>
> I don't know how to fix this.
>
> Kris

The MOVED data base is usless from my point of veiw. Only useful purpose
it serves is to explain what happened to a port that is no longer in the 
collection and it does that poorly.  Seems like CVS records would be a better 
way to follow a port's history.

IMO a port should only have one entry, and that is only if it was removed, 
here is how the fields might be set up:

category/portname | date removed | reason

If the port was renamed in the reason column just enter something like:

renamed: new category/ portname

If the port is placed back into service then remove it's record.

Just my opinion...

-Mike