Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 27 Sep 1995 11:09:05 -0700 (MST)
From:      Terry Lambert <terry@lambert.org>
To:        patl@asimov.volant.org
Cc:        gryphon@healer.com, terry@lambert.org, hackers@freebsd.org, jmb@kryten.atinc.com, peter@taronga.com
Subject:   Re: ports startup scripts
Message-ID:  <199509271809.LAA10620@phaeton.artisoft.com>
In-Reply-To: <9509271454.AA02380@asimov.volant.org> from "patl@asimov.volant.org" at Sep 27, 95 07:54:45 am

next in thread | previous in thread | raw e-mail | index | archive | help
> I think I missed some of the details of the Makefile proposal.  How
> does a package install it's make dependancies?  Does it add lines to
> the makefile, or is the makefile dynamically generated by concatenating
> a bunch of small per-service make fragments?

I like the ability to manage a dependency graph, but I also think that
the makefile approach is overly complicated... mostly because of the
shared binaries that would have to be invoked before /usr is mounted
given the current arrangement of shared library files (not that I agree
with that arrangement, or that it shouldn't change to allow a purely shared
binary installation).

I think my biggest objection is that it requires install-time configuration
administration as part of the install.  I'd like to divorce that.  Say
we have a big OEM (like Rod 8-)) who puts BSD on every box.  He will put
defaults for the configuration he sells in the packages themselves, and
forcing install time configuration will be largely redundant (and very
frustrating for Rod).

> If it adds lines to the makefile, it still suffers from the auto-edit
> safty concerns that the straight control file scheme does.  (Although
> it does eliminate the concerns about getting the order right.)

It doesn't.  Concatenation is the method I'd see it using.

> If it includes a bunch of small fragments, that seems like an unnecessary
> extra complication.

So does adding a select(2) system call when you have a perfectly good
pipe(2) system call. Sometimes adding complexity to get features is a
good thing.  8-).

> And I think that either the file-name-ordering or straight-control-file
> scheme would make it easier to quickly review which services are started
> in what order than any make-based scheme.  (But I'm willing to be convinced
> otherwise.)

I definitely agree here.


					Terry Lambert
					terry@lambert.org
---
Any opinions in this posting are my own and not those of my present
or previous employers.



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199509271809.LAA10620>