sysutils/cfs

[perryh at pluto.rain.com]

Doug Barton <dougb at freebsd.org> wrote:
> On 09/07/2011 00:07, perryh at pluto.rain.com wrote:
> > Doug Barton <dougb at freebsd.org> wrote:
> >>>>>>> Better to deprecate such non urgent ports, & wait a while
> >>>>>>> after next release is rolled, to give release users a warning
> >>>>>>> & some time to volunteer ...
> >>>>
> >>>> That's an interesting idea, but incredibly unlikely to happen.
> >>>
> >>> It _certainly_ won't happen if those in charge refuse to try it!
> >>
> >> My point was that the idea is impractical ...
> > 
> > How is it impractical to, as a rule, set an expiration date based
> > on an anticipated future release date rather than only a month or
> > two out from when the decision is made? 
>
> As has repeatedly been explained to you ...

I think you may have gotten me confused with someone else.

> you're asking the wrong question. The question is, how does it
> benefit the users to leave it in when we know that we're going
> to delete it?  Either way the user will discover that the port
> is not easily available for installation when they update their
> ports tree.

Reread the first paragraph.  Provided the port is still in the
tree, when they try to build it the ports mechanism reports the
FORBIDDEN/BROKEN/whatever which describes the problem, and the
expiration date a month or two out.  (If the expiration date is
not included in the report, it should be.)  They then know that
they need to fix the port, or find someone to fix it, and they
know _why_ it needs to be fixed.  In contrast, if the port is
_no longer_ in the tree, they have no clue why it disappeared.

> The difference is that in the meantime people doing work on
> the ports tree don't have to work around the old port (that's
> going to be removed anyway).

It's only going to be removed if no one fixes it.  The whole
point is that "release users" don't continuously monitor their
ports -- they only upgrade when they become aware that they
need to (e.g. when a newer release becomes available).  The
proposal is to increase the liklihood that, come upgrade time,
a "release user" gets a specific, actionable description of
any problems that have arisen, rather than having a port that
they have been using mysteriously disappear.

> The point has repeatedly been made that with almost 23,000
> ports in the tree both innovation and maintenance become
> significantly more difficult. Keeping that burden as low
> as possible is a feature.

s/point/claim/

Last I checked, freebsd.org was claiming that the very large number
of supported ports was a feature.  It seems that some of the ports
committers disagree.

> To answer your question more directly, start thinking through all
> the possible permutations of having 4 completely separate branches
> of FreeBSD active and supported at the same time, with releases
> happening several times a year.

So define a variable along the lines of NEXT_PORT_PURGE_DATE
in one of the /usr/share/mk/bsd.port*.mk files, which (being
part of the base) _are_ branched, and when a situation of the
kind under discussion arises set the port's EXPIRATION_DATE
to NEXT_PORT_PURGE_DATE.  (And before you start jumping all
over the details, I recognize that this is a rough first cut
at a mechanism that would need some details fleshed out.)

> Now consider how to handle EOL branches.

Last I heard, the ports tree does not claim to support EOL
branches.  Why does this proposal need to?

> Then consider that FreeBSD has _never_ supported a release-branched
> ports tree, precisely because it's a huge amount of additional work
> that we don't have person-hours for.

How does this proposal require a release-branched ports tree?

> I realize that what you're proposing sounds attractive from a
> purely theoretical standpoint. The problem is that it increases
> the maintenance burden a non-trivial amount without providing
> any substantive benefit.

Benefit:  see above.  Maintenance:  I would not think that updating
NEXT_PORT_PURGE_DATE as required -- a couple of times per release
cycle, maybe a few more if the schedule slips repeatedly -- would
constitute a significant additional maintenance burden.