From owner-freebsd-current Tue Apr 30 12:55:47 2002 Delivered-To: freebsd-current@freebsd.org Received: from dan-gm.gm.dan.co.jp (dan-gm.gm.dan.co.jp [61.194.1.58]) by hub.freebsd.org (Postfix) with SMTP id 5562237B41A for ; Tue, 30 Apr 2002 12:55:39 -0700 (PDT) Received: (qmail 60776 invoked from network); 30 Apr 2002 19:55:34 -0000 Received: from h196.p970.iij4u.or.jp (HELO dan-pbg4) (210.138.202.196) by dan-gm.gm.dan.co.jp with SMTP; 30 Apr 2002 19:55:34 -0000 Date: Wed, 1 May 2002 04:55:25 +0900 Subject: Re: Save a few hunderd kilobytes or a few hundred perl users? Content-Type: text/plain; charset=US-ASCII; format=flowed Mime-Version: 1.0 (Apple Message framework v481) Cc: freebsd-current@FreeBSD.ORG, perl5-porters@perl.org To: Mark Murray From: Dan Kogai In-Reply-To: <200204301803.g3UI30oI031200@grimreaper.grondar.org> Message-Id: <36DA0290-5C74-11D6-84FE-00039301D480@dan.co.jp> Content-Transfer-Encoding: 7bit X-Mailer: Apple Mail (2.481) Sender: owner-freebsd-current@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG First my apology for choices of stronger words than they have to. On Wednesday, May 1, 2002, at 03:03 , Mark Murray wrote: > Well, look at it this way. Perl is very hard to build already, and it > is very big, and you say it is getting bigger. All "base" freebsd needs > is the core language. The rest bloats the source tree, slows down builds > of the whole operating system and provides copious opportunities for > cross-builds and upgrades to fail. One of the reasons I have chosen FreeBSD over Linuxen is its tidiness and slimness. So I do understand your concerns. > Can we not come to a compromise here? One possible solution might be as follow; rename /usr/src/contrib/perl5 to /usr/src/contrib/miniperl5 and just add enough file to build miniperl. miniperl it may be it has all functionalities that should be required to 'make world' -- that is, of course, unless the build process uses external module. I don't think anyone would object to that (AFAIK you need perl to build kernel). > And it sounds like a perfect candidate for a FreeBSD "port". The > bash(1) developer develops on FreeBSD (IIRC). That is a port. I have > no idea how many other of our 6000+ ports are developed on FreeBSD, > we dont have those in the base system unless they are needed for > the core operating system (such needs are things like, say, > OpenSSH, Kerberos5/Heimdal, Bind, Less, etc in src/contrib/). Yes, I want p5-* cleaned up as well. I just 'grep ^p5- /usr/ports/INDEX' and found 660! That's way too many (or, at least two of which I have developed :). Maybe we should BSDPANize all these. I think if whole perl5 is distributed via ports only it wouldn've raised this much rants. FreeBSD does need perl in core but not the whole thing and that is the problem. But rule of the thumb is NOT TO REMOVE THE SOURCE. /usr/src MAY BLOAT because it doesn't get installed by default (there are already bloated charmingly; 34853 files under /usr/src on FreeBSD 4.5-stable). If you need size control do so via install process. I would love to help in this area. I think just a little tweak to hints file and you should be able to build a minimum perl just by passing right Configure directive. > You realise that you are asking for FreeBSD to bloat itself to > unusable levels by setting this precedent? How many _other_ > modules are coming in? How big is Perl going to get? How much > longer is it going to take to build? What other software authors > will thus have valid reasons for having _their_ software as part > of the base system instead of as a port? I don't mean to include those that don't come with perl-x.x.x.tar.gz. CGI.pm doesn't look like a necessity and it may even be true. But the perl community voted to include it to perl standard distribution. Besides, this 'pick, grok and trash or keep' process should be too time-consuming. perl-current is already > 3000 files big and it is already beyond a power of individual to look through every one of them. You need a better approach than handpick CGI.pm and others. >> P.S. I would rather choose the NetBSD way of detaching Perl from core >> distribution altogether. That is far more politically correct. > > There is merit to this point - make Perl5 a "super-port" (or > something), that is closer to the OS than a usual port but not part > of the base OS. I have no objection to this. Maybe python and ruby should go for that approach as well and I see that's the way to go -- for ports. We still need perl to build FreeBSD and we got to come up with a correct soultion -- not only politically but also technically. Your current soultion is, to say the least yet with all due respect, incorrect in both criteria. Dan the Proud Member of Both Communities To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message