Subject: Re: Release cycles (Was: Re: RealAudio)
To: Jim Bernard <>
From: John F. Woods <>
List: current-users
Date: 11/01/1997 08:45:51
> > Hopefully, the 1.4 release cycle will not last nearly as long as 1.3 :-)
>   As I recall, this was said of 1.2 and 1.3 as well.

Actually, I think the 1.2->1.3 gap *was* less than 1.1->1.2.

>  If I may make a somewhat radical suggestion: why not simply dispense
> with releases altogether?  Instead simply let portmasters (or other
> interested parties) put up snapshots (with source, please!) whenever the
> spirit moves and the tree is not too badly broken.  Anyone who wants to
> put out a CD can simply gather up the most recent snapshots from each of
> the ports, perhaps with -current source thrown in for do-it-yourselfers.

In a sense, that's what the Linux world tends to do; each major CD vendor
tends to release CDs on their own schedule with their own favorite fixes,
and are only slightly influenced by each other and by whatever schedule
Linus has.  The difficulty is support:  when someone complains that their
NetBSD is broken, if they say they're running "1.2", you know off the bat
(well, eventually you know) what bugs they are known to have, so in most
cases you can give them a quick answer.  (Of course, when the answer is
always "get -current", you introduce them to Constant Upgrade Hell.)  With
the snapshot-per-port scheme, each portmaster ends up being the primary
support point simply because they're likely the only ones who know for sure
what went into each snapshot.  Any portmaster without the resources of a
sizable company is likely to get buried by more than one or two such
snapshots per year.  Either that, or the answer will, again, be "get
-current", in which case anyone who either doesn't want to compile their
own system or who simply doesn't have time to do so will instead switch
to FreeBSD or Linux.  (And for the usual suspects who express utter shock
that *anyone* would not be *eager* to recompile their whole system on a
moment's notice:  there's a boatload of stuff I'd be able to compile
*instead* if I didn't spend so much time chasing -current.  It's all just
a hobby for me, so it doesn't matter to me, but *some* people actually
buy these expensive computer thingies to (you'd better sit down here) get
useful work out of them.)

There are also other problems of practicality:  the tree is *often* broken
for some ports and not others, so without a concerted effort by the core
group you *can't* get simultaneous snapshots (and if the timing is solely
determined by the portmasters' day-to-day mood, you *won't* get them, either);
this means that someone intending to put out a multi-architecture CD has to
have a separate source tree for each architecture -- this is not the way to
encourage companies to package NetBSD on CD.