Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 24 Jan 2000 12:16:24 +0800
From:      Greg Lehey <grog@lemis.com>
To:        Andrzej Bialecki <abial@webgiro.com>
Cc:        Andrew Hannam <famzon@bigfoot.com>, small@FreeBSD.ORG, jkh@FreeBSD.ORG
Subject:   Re: One disk vs Two Disk (was Re: New approach to picobsd)
Message-ID:  <20000124121624.C2643@mojave.worldwide.lemis.com>
In-Reply-To: <Pine.BSF.4.20.0001232335180.43769-100000@mx.webgiro.com>; from abial@webgiro.com on Mon, Jan 24, 2000 at 12:09:43AM %2B0100
References:  <005001bf6588$09fb8760$0104010a@famzon.com.au> <Pine.BSF.4.20.0001232335180.43769-100000@mx.webgiro.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Monday, 24 January 2000 at  0:09:43 +0100, Andrzej Bialecki wrote:
> On Sun, 23 Jan 2000, Andrew Hannam wrote:
>
>>> Do we need to do it in one disk?
>> The answer is a resounding "YES" for as long as possible.
>
> [lots of good suggestions]
>
>> With this problem solved I was able to remove such things as ee and more.
>> The command line then becomes purely an "advanced" user fixit system.
>
> A couple of general thoughts:
>
> * current method of building: the way we do it right now, with crunchgen,
> is relatively simple as long as you use system programs with well-formed
> Makefiles and consistent build process. But even here there are a couple
> of tough cases,

Indeed.  I don't know how easy it would be to include ports in the
framework, for example.

> to say nothing about what happens if you try to incorporate some
> external utility. Also, this method doesn't allow for any modularity
> (adding/removing programs without the need to rebuild everything
> from sources).

I don't understand this.  You need to modify crunch.conf, but once you
have done that, you just run make again, and it only builds the
necessary files.

> But this requires some clear path for how to build external utilities
> (e.g. from ports) so that they work in this model.
>
> I don't have any clear picture how to do this. Last time I tried using
> dynamic linking instead of static, it took so much more space that I gave
> up immediately. Also, I didn't know how to minimize the set of symbols
> from standard shared libraries to contain only those referenced by loaded
> programs - if you don't do it, you need to use full standard libs which
> are _big_.

Is that correct?  I thought crunch only included those library
functions which were actually referenced, directly or indirectly, by
the objects you included.  But I haven't looked at the process that
carefully.

> * the step that we can already take is to make use of kernel modules,
> instead of compiling everything into the kernel statically. Many ethernet
> drivers can now be loaded on demand. We could include them on another,
> "modules" floppy, freeing some space from the base floppies which would
> contain the smallest subset of drivers possible. Then, you could download
> the "base" floppy, "drivers" and combine them together to match your
> hardware, without recompiling the whole stuff (which I see as a big
> obstacle in using picobsd).

This is a nice idea conceptually, but like with dynamic linking, the
result is bound to be bigger.

> BTW, I think this is something to be considered for installation
> floppy as well. Jordan, are you listening?

Agreed.

> * incorporating more small, useful tools: recently I ported SASH which
> will be great addition for setups like "router" floppy.

I think SASH might have great potential in the "first or only" disk.
It would be really nice to have a single disk which could be extended
by further disks if you wanted.  That would pave the way to a single
"do it all" PicoBSD approach.

Greg
--
Finger grog@lemis.com for PGP public key
See complete headers for address and phone numbers


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-small" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20000124121624.C2643>