cvs commit: ports/misc/cwish Makefile

[Mark Linimon]

On Tue, 13 Apr 2004, Kris Kennaway wrote:

> BTW, when removing the last master site, you should also mark the
> port BROKEN if no copy of the distfile exists on ftp.freebsd.org.

A good point, but I have been doing a 'make fetch' first.  (I did
notice one that you had marked broken/unfetchable that did indeed
have copies of the distfile on ftp.freebsd.org -- should I mark that
one as unbroken?)

But your response gives me the idea to ask a related question ...

When I proposed the DEPRECATED makevar, my intention was to use it to
be the union of (BROKEN or FORBIDDEN or "replaced by a newer version").
So far IMHO it's serving that role well.

But in going through the distfile survey for unfetchable/unmaintained
ports (which is what I've been doing), I am finding a number of ports
like this where the original developer's site has vanished and a web
search shows that there has been nothing to replace it for months or
even years.  (In fact, doing a web search produces only "ghosts":
references to CVS commit logs, other postings to mailing lists, the
port's status page on FreshPorts, and sometimes other, similar, pages
for one of (NetBSD|OpenBSD|Debian)).  

Now, in many of these cases, the code still runs, so I don't see any
particular need to delete them ... if someone is using them.

But should it be fair game to mark them as DEPRECATED with the text
"development seems to have ceased" (or equivalent)?

This is something that we would need to see if there is (or can be)
a community consensus on.  On the one hand, if they seem to install,
and have no PRs against them, they are "useful" (for some definition
of "useful").  On the other hand, if there are alteratives available,
shouldn't we be directing prospective users to the alternatives, and
even if there aren't, who's going to bother to fix them when they
break?

btw, I'm not going to take any action on this one way or the other,
pending any particular consensus -- I'm just interested in bringing
up the point.

A final datapoint: my guess is that we have at least 200 ports like
this.  With a little bit of work, I could add such a report to portsmon,
or write it as some kind of standalone sh-based code.

mcl