Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 03 Aug 1997 07:33:39 +1000
From:      David Nugent <davidn@unique.usn.blaze.net.au>
To:        "Jordan K. Hubbard" <jkh@time.cdrom.com>
Cc:        Michael Smith <msmith@atrad.adelaide.edu.au>, asami@cs.berkeley.edu (Satoshi Asami), andreas@klemm.gtn.com, ports@FreeBSD.ORG, current@FreeBSD.ORG, stable@FreeBSD.ORG
Subject:   Re: ports-current/packages-current discontinued 
Message-ID:  <199708022133.HAA14498@unique.usn.blaze.net.au>
In-Reply-To: Your message of "Sat, 02 Aug 1997 11:32:38 MST." <15386.870546758@time.cdrom.com> 

next in thread | previous in thread | raw e-mail | index | archive | help
>> I have seen in the FreeBSD project since my involvement. I can count the
>> number of such incidents I've witnessed in the last couple of years on one
>> hand, so it's not like the project is infested with stupidity. It was
>> very ill-considered, and Satoshi's position here is critical. That he
>> apparently got no say in this is incredible, to say the least. It seems
>> obvious to an outsider that there are some very fundamental communication
>> problems within the core team. Seeing this fixed is even more critical
>> than where tcl or perl happen to reside.
>
>What most people here don't understand is that there was very fierce
>debate about this behind the scenes for some time before it was done,
>and what started as a TCL debate blew up into a whole "should FreeBSD
>have everything unbundled, from the compilers to perl, or should it
>bundle all the high level tools so that other system tools can be
>written which depend on them?" sort of fracas.  One man's useful tool
>is another man's wasteful, unnecessary bloat, it seems, and rather
[..]

Yes, yes, yes. But all completely besides the point.

I don't care one way or another about "bloat". It is irrelevent
to the point I was making. "Bloat" is subjective, as you point out,
but it has nothing to do with how tools are updated and integrated
into the system. Tools which are imported into the base system are
*necessarily* more static, because of the very real concerns that
have caused problems here. They won't change as often, they can't
change as often, else we should by rights be carefully following
perl development IN THE SOURCE TREE rather than having not only an
obsolete version of perl, one completely unsupported and with
seriously security bugs, and one which hardly anyone uses anyway.
Tcl, until this apparently unpopular import, was at least in a
better position than that.

I've said before that perl4 must die. I don't particularly care
whether it is imported into the source tree or left as a port.
It is stale, it is old, it is obsolete and completely unsupported.
It has bugs, limitations and security holes. It must go. Period.
My only concern about importing a current version into the source
tree is the same as tcl - what guarantees are there that this will
be kept up to date?

The ports system, in spite of its faults and shortfalls, *works*,
and works extremely well, mainly because of the efforts of porters
and Satoshi's management. But I hardly need to point this out to
you of all people. The ports system integrates third party software
into the base system seemlessly, provides upgrade paths and has
people actively supporting (the most popular packages, anyway) and
updating their favourite apps. Sure, there are mistakes there, like
the recent apache vs. apache current debate, but these are far more
easily handled in the less expensive and more low-maintenance system
that ports has been designed to handle.

Again, this has nothing to do with any "bloatist" point of view,
nor is it anti-perl or anti-tcl either - I use both daily and
enhance systems constantly with both of them.

BUT:

How is tcl essential to a working FreeBSD environment? How is
perl, for that matter, other than a few scattered scripts which
can and do work quite happily with current versions of perl
(and can easily be replaced with /bin/sh scripts anyway)?
Why are we nailing ourselves to versions of third party
software which are prone to become stale and unmaintained?
Why can't the ports system be used for them?

>Once you start with TCL, it will *not* stop there - I can only assure
>you of that.  Perl will follow immediately behind, as will much other
>stuff (yes perl fans, there are many out there who consider your
>favorite utility language an evil, bloated monster which should not be
>bundled with FreeBSD at all).  What we have now is a rough state of

Surely, you're not suggesting that just because something is
in not in the FreeBSD base system that it is "evil"? Oh come on,
this is such a childish attitude I can't believe I'm hearing it.
The question is not how large a package is, it is how useful it
is in the base distribution - how dependant upon it a FreeBSD
system is in installation and setup, and how integrated it is
into the system. Neither perl nor tcl have any grip in the system
where they could be considered so essential that FreeBSD will not
run without them. It is more *appropriate* that they be maintained,
integrated and - ESPECIALLY - kept current via the ports system.

This is a major PLUS for users, not the opposite as you seem to
be assuming. perl4 in the tree is living proof of exactly where
we end up with the approach that is being taken.


>equilibrium between the two sides (who are fundamentally at odds as to
>what constitutes a reasonable bundling policy) and while TCL may be
>causing some grief, I think the bloatists are content with that state
>of affairs and nothing else is on the bundling horizon that I can see.
>Nuke TCL and you will swing the balance in the other direction, with a
>lot more than just TCL biting the dust as a result.  Maybe that's not
>such a bad thing, but just so you understand how much of a "linchpin"
>issue this one is.
 
No, I understand the issue well enough. But I think it is
irrevelent. Having "current" versions of things in ports and
"old" things in the tree is not only annoying, it is downright
ridiculous. The opposite is almost as bad. I would quickly add
that any generalisation here is dangerous - you have to
consider the benefits of things being integrated into the
source tree vs. the costs of possibly becoming stale. It is
far easier to keep our cvs "hands" of a third party package
and integrate it via the ports system. Diffs between a
distribution and the "FreeBSD version" are absolutely plain
without even having a source repository around. Things like
bind, gcc, gdb etc. are a completely different ballgame.

So, no, I don't see this as any sort of weighted argument. You
can cry about bloat all day, and I simple don't care. That
point of view misses the issue entirely. tcl is a major headache
in terms of multiple version operability. Perl is likewise
(do you do much perl debugging?). I'd live silently with these
problems if the benefits outweighed the cost in terms of hassle,
but to me, it just isn't worth it.


Regards,
David

David Nugent - Unique Computing Pty Ltd - Melbourne, Australia
davidn@freebsd.org davidn@blaze.net.au http://www.blaze.net.au/~davidn/



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199708022133.HAA14498>