[9fans] x10

[9fans] x10

[Fco.J.Ballesteros]

I've pushed to sources a program called lights,
that is a gui for the x10 file system. It can be used to
see the status of lights/sensors and to change them by
using the mouse.

The file system (cm11 driver) has some bugs fixed too.

hth

Re: [9fans] x10

[Dave Lukes]

Damn!

All I need is a place to live, a load of X10 stuff and a plan9
system and I can really start to have some fun:-~.

Dave.

On Thu, 2004-04-01 at 09:45, Fco.J.Ballesteros wrote:
> I've pushed to sources a program called lights,
> that is a gui for the x10 file system. It can be used to
> see the status of lights/sensors and to change them by
> using the mouse.
> 
> The file system (cm11 driver) has some bugs fixed too.
> 
> hth
> 
> 

Re: [9fans] x10

[boyd, rounin]

> All I need is a place to live, a load of X10 stuff and a plan9
> system and I can really start to have some fun:-~.

all i need is a large amount of money [150k EUR] and then ...

Re: [9fans] x10

[Kenji Okamoto]

> all i need is a large amount of money [150k EUR] and then ...

By the way, it seems that the free software movement may have
shifted to Europe these days.   I got region 1 "Gangs of New york",
and had to setup my Linux box to enjoy it.   The ogle movie free
software is from se domain...   What is happening in US these days.

Kenji

Re: [9fans] x10

[boyd, rounin]

> The ogle movie free software is from se domain...

under .se law that is 'fair use' [copy whatever you like, but don't use it
for a commercial purpose].

Re: [9fans] x10

[a]

// ...it seems that the free software movement may have shifted to
// Europe these days. ...What is happening in US these days.

all americans are either cowering in their duct-tape-sealed homes
or have left the country. besides, only terrorists use free software.

(yes, it's meant to be a joke, although i wish it was farther from
reality. bah.)

also: bomb, terror, semtex, NSA, congres. hello eschelon, carnivore!
ア

Re: [9fans] x10

[a]

// Just look at the world of Windows. Where is all the
// BSD licensed Windows software?

um, there's not exactly a *ton* of GPL code for Win32, either. some, sure,
but the dearth of both is due, i believe, do the nature of the underlying
system. 

// With no GPL I'm willing to bet that plan9 would not have gone open source...

while i'm not willing to take that bet, i'm not backing it, either. several of
the labs folks have been fighting with lawyers to get as much out the door as
possible for years before the GPL was the force it is today. and i believe the
shift in the lawyers ability to listen is due more to changes in the world at
large than to the GPL (although it likely helped).

that said, you point is probably valid: it's certainly been the case with
various hardware manufacturers that produce drivers for linux that wouldn't
otherwise be likely to give out the info.

anyway, i'm not looking to put down the GPL. it's fine for what it is and
what it does. it's actually *especially* useful for companies producing their
own code, all FUD aside. what strikes me as "bad" is the idea that the GPL
has a moral advantage over BSD/MIT licenses.

anyway, tangential references to plan9 aside, this has gone fairly far off
topic. whatever license you like is fine: if i don't like it, i simply
won't use your stuff. there's always a choice.
ア

RE: [9fans] x10

[ron minnich]

On Tue, 6 Apr 2004, Tiit Lankots wrote:

> > Actually there's another reason to choose Linux last (other than
> > Windows): the distributions I've seen are just a collection of ~1,000
> 
> Just for kicks, I checked the mozilla binary on my Slackware. On startup,
> it has to drag in 29 (!) DLLs, some of which need other DLLs themselves.
> 'Orrible.

yeah, well, just watch date(1) resolve 213 symbols sometime. to print
today's date. Including threaded posix libraries for dns. It's really
amazing, after we just saw Linus argue on this list how important TLB
fault minimization is, to see how inefficient so many things are on Linux.
They are so microbenchmark-driven that they oft-times miss the big
picture, I think. 

ron

Re: [9fans] x10

[Derek Fawcus]

On Tue, Apr 06, 2004 at 12:03:28PM -0600, ron minnich wrote:
> 
> yeah, well, just watch date(1) resolve 213 symbols sometime. to print
> today's date. Including threaded posix libraries for dns.

Yeah it's a pain in the arse.  I tend to compile things static these days -
using a dietlibc.

> It's really
> amazing, after we just saw Linus argue on this list how important TLB
> fault minimization is, to see how inefficient so many things are on Linux.
> They are so microbenchmark-driven that they oft-times miss the big
> picture, I think. 

I suspect a lot of people ignore the costs and delays being added to user
mode...

After severly hacking a RH install I got a more or less comfortable
environment on my laptop,  and a not unreasonable boot time - so I
never suspend,  just shutdown... but the laptop'll be replaced soon,
so time to battle again - that'll be when plan9 gets a tryout on it.

DF

Re: [9fans] x10

[ron minnich]

On Tue, 6 Apr 2004, Derek Fawcus wrote:

> After severly hacking a RH install I got a more or less comfortable
> environment on my laptop, and a not unreasonable boot time - so I never
> suspend, just shutdown... but the laptop'll be replaced soon, so time to
> battle again - that'll be when plan9 gets a tryout on it.

two essential things Plan 9 lacks for laptops
- sleep that works
- hot plug devices (e.g. pcmcia, usb dongles, etc.)

ron

Re: [9fans] x10

[matt]

> two essential things Plan 9 lacks for laptops
> - sleep that works
> - hot plug devices (e.g. pcmcia, usb dongles, etc.)

usb hotswapping works for me

i just run usb/usbd again

m

Re: [9fans] x10

[boyd, rounin]

> yeah, well, just watch date(1) resolve 213 symbols sometime. to print
> today's date. Including threaded posix libraries for dns. It's really
> amazing, after we just saw Linus argue on this list how important TLB
> fault minimization is, to see how inefficient so many things are on Linux.

right.  run strace on any given linux binary and watch the torrent of
system calls before main() even gets called.  it is truly appalling:

RE: [9fans] x10

[rog]

> yeah, well, just watch date(1) resolve 213 symbols sometime. to print
> today's date.

it's funny, i was probably looking at the source to the plan 9 date
command as you were writing that...

it's perhaps an interesting example.  i was looking back in the dump
file system and wanted to know the exact modification times of some of
the files there.  ls changes its format for files older than 6 months,
so:

% cd /n/dump/2003/0916/usr/rog/limbo/schedule
% ls -l scheduler.b
--rw-r--r-- M 30 rog rog 10183 Sep 15  2003 scheduler.b
% cat /dev/time
 1081276575   1081276575299952521         8847783936768             305454592 % 
% fc 1081276575 6 30 24 60 60 x x x x -
1065724575
% ramfs
% echo -n ' 1065724575   1065724575299952521         8847783936768             305454592' > /tmp/time
% > /tmp/bintime; chmod 000 /tmp/bintime
% bind -b /tmp /dev
% date
Thu Oct  9 19:36:15 BST 2003
% ls -l scheduler.b
--rw-r--r-- M 30 rog rog 10183 Sep 15 22:07 scheduler.b
% 

and there i was, back in time 6 months, and the files display
their full timestamp as desired.

how pleasant it is to be using a system where almost all resources
are malleable, including time itself!

Re: [9fans] x10

[boyd, rounin]

> how pleasant it is to be using a system where almost all resources
> are malleable, including time itself!

nice one.

Re: [9fans] x10

[Enache Adrian]

On Tue, Apr 06, 2004 a.d., rog@vitanuova.com wrote:
> % bind -b /tmp /dev

That could be dangerous too.
I'd like a rfork flag to prevent bind() altogether.

Regards,
Adi

Re: [9fans] x10

[Russ Cox]

Enache Adrian wrote:
> On Tue, Apr 06, 2004 a.d., rog@vitanuova.com wrote:
> 
>>% bind -b /tmp /dev
> 
> 
> That could be dangerous too.
> I'd like a rfork flag to prevent bind() altogether.

why?  you can only hurt yourself.

Re: [9fans] x10

[Enache Adrian]

On Tue, Apr 06, 2004 a.d., Russ Cox wrote:
> Enache Adrian wrote:
> >On Tue, Apr 06, 2004 a.d., rog@vitanuova.com wrote:
> >
> >>% bind -b /tmp /dev
> >
> >That could be dangerous too.
> >I'd like a rfork flag to prevent bind() altogether.
> 
> why?  you can only hurt yourself.

How can I enforce static namespace semantics upon a process ?
Stop it from lying to its children about /dev/zero being /dev/random ?
Or letting them happily fill up the disk when writing to /dev/null ?

Regards,
Adi

Re: [9fans] x10

[Geoff Collyer]

I don't think that these are actual security problems.  Remember that
Plan 9 doesn't have set-id, so if you fool one of your child
processes, you can only hurt yourself.

Why do you want to enforce a static namespace?

Why would you bind /dev/random onto /dev/zero and why would you
consider it a security problem if you did?  Likewise for binding an
ordinary file onto /dev/null.

Plan 9, like Unix, will let you shoot yourself in the foot if that
makes you happy.  But shooting yourself in the foot is not a security
problem.

Re: [9fans] x10

[Enache Adrian]

On Wed, Apr 07, 2004 a.d., Geoff Collyer wrote:
> Why would you bind /dev/random onto /dev/zero and why would you
> consider it a security problem if you did?  Likewise for binding an
> ordinary file onto /dev/null.

What's wrong with enforcing upon a process truerand() returning
random numbers and not zero every time ?

bind() may dramatically change the meaning of any file operation on
Plan 9 - and is itself not constrained by file permissions, etc.
Sometimes one may want to turn such a powerful feature off :-)

And generally, being able to control with rfork() flags the couple
of things Plan 9 still do without reading or writing files
seemed to me like a not completely ireasonable idea.

Of course, that may not be as useful as I first thought.

Regards,
Adi

Re: [9fans] x10

[Geoff Collyer]

If you want to make sure that you're getting the real /dev/random,
unmount /dev/random (but not /dev) first, or bind -b '#c' /dev again.

I don't think anybody has been persuaded yet that it's worth expending
the effort and increasing the kernel's complexity to restrict bind
with rfork flags.

Re: [9fans] x10

[Russ Cox]

> What's wrong with enforcing upon a process truerand() returning
> random numbers and not zero every time ?

If you start the process yourself, then you'll know that /dev/random
is okay, assuming you're not shooting yourself in the foot on purpose.

If you start an untrusted program and then have it start the process,
all bets are off, as you have observed.  But if you're letting an untrusted
program start some other program, you have other things to worry 
about -- like the untrusted program might edit the binary to remove
all the calls to truerand() before executing it.  Or might just not run
the binary and run its own entirely different version.  There's no
sense in trying to lock down this one thing.  The solution is not to
rely on untrusted programs to do your work for you.

Russ

Re: [9fans] x10

[rog]

the reason this is a problem in unix is the setuid bit.  various setuid
programs expect to see files (e.g./etc/passwd) in their expected
place.  if you can replace 'em, you can break the system.  hence
chroot is a superuser-only system call.

if you don't have setuid programs, it's not a problem.

as far as i can see, this was one of the major "too deep to fix"
problems with unix.

Re: [9fans] x10

[Charles Forsyth]

>>about -- like the untrusted program might edit the binary to remove
>>all the calls to truerand() before executing it.

for instance, and more easily, by dynamically substituting an unexpected libc-2.74.5.beta-9

Re: [9fans] x10

[Bruce Ellis]

do they have glibc swap meetings there?  i hear they are popular with
the pre-teens in japan.

brucee
----- Original Message ----- 
From: "Charles Forsyth" <forsyth@terzarima.net>
To: <9fans@cse.psu.edu>
Sent: Friday, April 09, 2004 8:09 AM
Subject: Re: [9fans] x10


> >>about -- like the untrusted program might edit the binary to remove
> >>all the calls to truerand() before executing it.
> 
> for instance, and more easily, by dynamically substituting an unexpected libc-2.74.5.beta-9
> 
> 

Re: [9fans] x10

[lucio]

> How can I enforce static namespace semantics upon a process ?
> Stop it from lying to its children about /dev/zero being /dev/random ?
> Or letting them happily fill up the disk when writing to /dev/null ?

By changing #/, of course.  Just mount whatever root server you see fit.

++L

Re: [9fans] x10

[Dave Lukes]

> That could be dangerous too.

How?

> I'd like a rfork flag to prevent bind() altogether.

Why?

	Dave.

Re: [9fans] x10

[matt]

>> I'd like a rfork flag to prevent bind() altogether.

> Why?

if someone supplies you a binary you could fork a shell without binding, clear the namespace and run the supplied binary. Like chroot & jailing it

However, preventing binds altogether seems a bit strange

if you fork with 

          RFNOMNT   If set, subsequent mounts into the new name space
                    and dereferencing of pathnames starting with # are
                    disallowed.
and

     RFCNAMEG  If set, the new process starts with a clean name
                    space. A new name space must be built from a mount
                    of an open file descriptor.  Is mutually exclusive
                    with RFNAMEG.

what can you lose from allowing the binary from manipulating it's own namespace?

m

Re: [9fans] x10

[Martin C.Atkins]

On Tue, 6 Apr 2004 03:41:40 -0700 9nut@9netics.com wrote:
> By definition very few have seen the closed ones, and can't really
> talk about it.  Five will get you ten that Windows XP source is orders
> of magnitude messier that Linux.

Yes. I'd place odds too!

A related issue: I read somewhere that the source code for accessing
outlook .pst mail boxes has been completely lost. All that now exists
is various .dll libraries that can't be re-built!

Martin

-- 
Martin C. Atkins			martin@parvat.com
Parvat Infotech Private Limited		http://www.parvat.com{/,/martin}

Re: [9fans] x10

[Geoff Collyer]

Sounds like there's a business opportunity to sell Microsoft some Ken
file servers with optical jukeboxes or fossil/venti servers with RAID
disks.  Maybe a few CD-RW and DVD-RAM drives too.

Or maybe they just can't *find* the sources.  A business opportunity
for Google.

Re: [9fans] x10

[9nut]

> Sounds like there's a business opportunity to sell Microsoft some Ken
> file servers with optical jukeboxes or fossil/venti servers with RAID
> disks.  Maybe a few CD-RW and DVD-RAM drives too.
> 
> Or maybe they just can't *find* the sources.  A business opportunity
> for Google.

I was going to say that maybe finding sources to dll's they can't rebuild
would be a good test case for their (new to them) search engine.

Re: [9fans] x10

[boyd, rounin]

> I was going to say that maybe finding sources to dll's they can't rebuild
> would be a good test case for their (new to them) search engine.

maybe they've lost the expertice needed to code them.  a windoze DLL is a
scary thing and it's even worse when you have to code one.

Re: [9fans] x10

[boyd, rounin]

> ... or fossil/venti servers with RAID disks.

i had this conversation with a mate of mine the other week.

RAID controllers are expensive, overly complex and prone
to failure.  i didn't agree 'cos i'd seen one that didn't fail.

however i was compelled by his logic that the controller
i was using [Digital Storageworks] expensive and
probably over complex.

we agreed that mirroring is simpler, less error prone and
now that disks are cheap going to RAID5 is a complete
waste when it can be trivially replaced with mirroring.

Re: [9fans] x10

[Martin C.Atkins]

On Wed, 7 Apr 2004 15:36:04 +0200 "boyd, rounin" <boyd@insultant.net> wrote:
> > ... or fossil/venti servers with RAID disks.
> 
> i had this conversation with a mate of mine the other week.
>...
> we agreed that mirroring is simpler, less error prone and
> now that disks are cheap going to RAID5 is a complete
> waste when it can be trivially replaced with mirroring.

This link, from the thread "Rob makes Dave Farbers 'vanity publishing' list":
	http://www.interesting-people.org/archives/interesting-people/200404/msg00066.html

also makes the argument that in a very big deployment - they have
100,000 servers, each with 2 disks (yes, I counted the zeros!) - RAID
doesn't make much sense. I was convinced.

Martin

-- 
Martin C. Atkins			martin@parvat.com
Parvat Infotech Private Limited		http://www.parvat.com{/,/martin}

Re: [9fans] x10

[Geoff Collyer]

By `RAID', I just meant any form of redundancy so that data can be
recovered on-line when a disk blows.  I wasn't advocating any
particular RAID level.  The mirror device in Ken fs, for example, is
just fine.

Re: [9fans] x10

[Martin C.Atkins]

On Wed, 7 Apr 2004 21:45:34 -0700 Geoff Collyer <geoff@collyer.net> wrote:
> By `RAID', I just meant any form of redundancy so that data can be
> recovered on-line when a disk blows.  I wasn't advocating any
> particular RAID level.  The mirror device in Ken fs, for example, is
> just fine.
> 

Yes. I guess within the original meaning of the acronym, Google's scheme certainly
counts - rather nicely!

Martin
-- 
Martin C. Atkins			martin@parvat.com
Parvat Infotech Private Limited		http://www.parvat.com{/,/martin}

Re: [9fans] x10

[Geoff Collyer]

I hadn't considered that angle, I just meant that perhaps Microsoft are
so disorganised that they need a search engine to find their lost
source code.

Re: [9fans] x10

[Joel Salomon]

a@9srv.net said:
> > If every human being through the ages did that, [hid knowledge]
> > we'd all still be reinventing the wheel and throwing rocks.
>
> only in the absolute. allowing inventors to profit from their work
> also encourages future invention. there's a balance to be struck.
> current laws in the US involving generic drugs are one (very
> imperfect) example. original, pre-Disney copyright is another.
> ã‚¢
>

OTOH, patent law *does* require that the particulars of the invention be
released. You just can't *use* it w/o permission.

--Joel

Re: [9fans] x10

[Brantley Coile]

> a@9srv.net said:
>> > If every human being through the ages did that, [hid knowledge]
>> > we'd all still be reinventing the wheel and throwing rocks.
>>
>> only in the absolute. allowing inventors to profit from their work
>> also encourages future invention. there's a balance to be struck.
>> current laws in the US involving generic drugs are one (very
>> imperfect) example. original, pre-Disney copyright is another.
>> ã‚¢
>>
> 
> OTOH, patent law *does* require that the particulars of the invention be
> released. You just can't *use* it w/o permission.

And that is the crux of the matter.  You invent to do stuff help
create a product and sell it.  We (society) will give you a monopoly
on it for a very limited time if you tell us how you do it.  We will
stand on your shoulders and do new stuff based on what you have told
us.  Also, in 17 years we get to make more of them.

The original concept of the Patent was invented in England to
encourage Englishmen to figure out how the French were doing all the
fancy stuff they were doing in textiles.

Without doubt, if no patents, no Thomas Edison.  No Thomas Edison, no
power outlet.  No power outlet, all the computers would have to run off
kerosine.  : )

Brantley

Re: [9fans] x10

[matt]

"If Mr Edison thought a bit more, he wouldn't have to sweat so much."

No Mr Tesla, no AC power distribution

m

Re: [9fans] x10

[Brantley Coile]

> "If Mr Edison thought a bit more, he wouldn't have to sweat so much."
> 
> No Mr Tesla, no AC power distribution

Al fired Tesla.  Tesla never fired Al.  : )


(My inventor can beat up your inventor!!) : ) 

PS.  Besides, West Orange is close to Murry Hill.

Re: [9fans] x10

[Brantley Coile]

> "If Mr Edison thought a bit more, he wouldn't have to sweat so much."
> 
> No Mr Tesla, no AC power distribution

Al fired Tesla.  Tesla never fired Al.  : )


(My inventor can beat up your inventor!!) : ) 

PS.  Besides, West Orange is close to Murry Hill.

Re: [9fans] x10

[boyd, rounin]

> PS.  Besides, West Orange is close to Murry Hill.

hoboken train, waz the closin' doors.

Re: [9fans] x10

[Brantley Coile]

It's just too easy to hit that Post tag on acme!  Sorry for the double
post.

Did you notice that acme has the same letters as emacs?  I guess I'm
just dense and never noticed this before.  Missing the s though.  Of
course who would want to use an editor named sacme.  ...  or scame for
that matter.

 Brantley

Re: [9fans] x10

[boyd, rounin]

> No Mr Tesla, no AC power distribution

AC is good.

Re: [9fans] x10

[Kenji Okamoto]

Thank you, geoff.

In a short word, my argument is that we should pay more respects to those
who are sharing their own time to others.   Of course, I know we can employ
more people with high quality if more money.   This is true everywhere 
professional works are concerned though.   

It's easy to say that quality of some free software are bad from the view point 
of skilled professionals.   However, we must remind that the programmer who 
are working without pay are shareing their own time, not full-time of them, 
which is same to us after Plan 9 became to a free software.   As you know the 
quality of Plan 9 itself is now going down, if we cpmpare it with ,say, Release 2 
or 3.   I think this is very natural, because we have now no full-time programmer 
of Plan 9.

Being said above, I still love Plan 9 because of its beatiful design and codes,
very easy to understand and to read!   I feel Plan 9 has her own future, if we 
could continue to develope its philosophy.   The problem is how we can get 
more programmer even though we cannot pay them money in a future.   I think 
Linux people may have a solution or at least trying to get it with struggling.

>So my
> preferences in code would be, in order of decreasing desirability:
> Plan 9, Research Unix, OpenBSD (they at least try to pay attention to
> security issues), *BSD, Linux, full stop.

My preference is only Plan 9, because I don't know Research Unix.☺

> Actually there's another reason to choose Linux last (other than
> Windows): 

I chose Linux because it's the best spread lunix, and I can use it as my own
base for working (actually, I may not need sources of it, because it's just a
running base for applications to me).   From this view point, I doubt there is 
any value to pay attention to other Unices, because we have Plan 9.

> I don't know if that answers your questions, Kenji.

You did, geoff.

Kenji

Re: [9fans] x10

[Martin C.Atkins]

On Tue, 6 Apr 2004 00:11:23 -0700 Geoff Collyer <geoff@collyer.net> wrote:
>...

Lots of stuff I agree with...

> Actually there's another reason to choose Linux last (other than
> Windows): the distributions I've seen are just a collection of ~1,000
> packages and, if they even come with sources, you hope you can compile
> them all, but they demand different prerequisite library versions
> (often for little or no reason), some of which are incompatible and
> can't co-exist, so it becomes a major pain to construct a consistent
> source tree, for which you have running binaries, and which you could
> recompile all (or part) of.  The BSDs at least are integrated systems,
> maintained as systems, not packages, so they come with self-consistent
> sources, and you can compile them.  For programs that aren't part of
> the core system (``ports''), they let `make' figure out the
> dependencies (novel!) and drag in the necessary prerequisites and they
> usually build (occasionally the gratuitous overspecificity of version
> numbers bites one in the ass; it's a real problem in the BSD/Linux
> world).

This is also mostly true, but having used Debian Linux for several
years now, I can claim with some non-trivial experience that Debian is the
counter example (that proves the rule? :-).
Debian linux, like the BSDs also tries to release a 'system', not
just a collection of packages. That is why the Debian release cycle
is so long, and why the 'latest and greatest' version is always
rather out of date!

I apt-get modules daily (or at least, weekly), and have never had
one program/library break another, or any of the usual library-version hell.

Of course, if you step outside the packages in the stable system, then
things can get a little more complicated, but that is your choice, and
usually the package system still sorts things out with very minimal
hassle.

>..
> If you're bored and have an afternoon, a fast machine, a broadband
> connection and ~10GBs of disk to kill, ask FreeBSD to build `gnucash':
> it drags in code and compiles and drags and compiles, but the best
>...

Just last week I apt-getted a reasonably recent version of gnucash,
and that was all there was to it. Nothing broke, all the dependencies
were automatically updated, etc. etc. No scheme-version hell, either.
(It's just a pity that the up-to-date manual hadn't also been packaged,
but that only took a few minutes to fix).

The number of times you see instructions for the installation of packages
on RPM-based systems include the direction to "turn off package version
compatibility checking" is truly scary! I'm not even sure if apt has
such an option!

> And of course there are the people who insist on programming in
> (shudder) Perl.  Lordy, lordy, lordy.  One is reminded of Phil Wadler
> on XML (see /sys/games/lib/fortunes): it's not a language (at least in
> the sense of having a finite or at least well-defined grammar, or at
> least a grammar!, and some attempt at having a minimal, non-redundant,
> orthogonal set of constructs that cover the application space of
> interest), it's not a sharp tool but rather a swiss-army
> gasoline-powered chainsaw, and it's not particularly good for doing
> anything.  Sorry; I'm allergic to Perl.

Agreed! Agreed!!!!

Martin
-- 
Martin C. Atkins			martin@parvat.com
Parvat Infotech Private Limited		http://www.parvat.com{/,/martin}

Re: [9fans] x10

[Geoff Collyer]

Ah!  Now I understand why the users of Debian Linux are fanatical about
it.  I'd never heard a convincing explanation before but this makes a
lot of sense.  Should I ever have occasion to install Linux again,
I'll give it a try.

Re: [9fans] x10

[Martin C.Atkins]

On Wed, 7 Apr 2004 00:30:14 -0700 Geoff Collyer <geoff@collyer.net> wrote:
> Ah!  Now I understand why the users of Debian Linux are fanatical about
> it.  I'd never heard a convincing explanation before but this makes a
> lot of sense.  Should I ever have occasion to install Linux again,
> I'll give it a try.

I'm just hoping they can maintain the record, given that the
woody->sarge update will see the number of packages go from 3000 ->
10,000 (numbers off the top of my head, but you get the general
idea!).

Martin
PS
  Yam, apt for rpms, etc are trying to add the fundamental mechanisms
to other distributions, but without the (correct!) dependency meta-data in the
packages, that's all just pointless window-dressing. The meta-data takes
real effort by *all* the package maintainers, and that's not so easy
to retro-fit!

-- 
Martin C. Atkins			martin@parvat.com
Parvat Infotech Private Limited		http://www.parvat.com{/,/martin}

Re: [9fans] x10

[boyd, rounin]

debian do have a slow release cycle, which is a good thing.

however, i have seen sloppy apt-get's smashing things into
an unusable state.  this shows that there is something
fundementally wrong with this whole 'pull it across
the net' philosophy and this is before we get into
security issues.

Re: [9fans] x10

[a]

// however, i have seen sloppy apt-get's smashing things into
// an unusable state.  this shows that there is something
// fundementally wrong with this whole 'pull it across
// the net' philosophy and this is before we get into
// security issues.

no, i don't think it does. regardless of what model you use, 
package maintainers can always make mistakes, and there are
always security questions. the fact that the specific apt-get
implementation of the 'pull it across the net' model is 
flawed (but still pretty good) just means that apt-get needs
fixing.

how to do that without having dumps or snapshots is left as
an exercise for the reader. ;-)

the first time i cd'd into the BSD ports tree and did a make
that automatically pulled things over the net i was totally
blown away. even if i did have to go to dinner while the
package downloaded over my 14.4K modem. :-)
ア

Re: [9fans] x10

[Joel Salomon]

Geoff Collyer said:
> Actually there's another reason to choose Linux last (other than
> Windows): the distributions I've seen are just a collection of ~1,000
> packages and, if they even come with sources, you hope you can compile
> them all, but they demand different prerequisite library versions
> (often for little or no reason), some of which are incompatible and
> can't co-exist, so it becomes a major pain to construct a consistent
> source tree, for which you have running binaries, and which you could
> recompile all (or part) of.  The BSDs at least are integrated systems,
> maintained as systems, not packages, so they come with self-consistent
> sources, and you can compile them.  For programs that aren't part of
> the core system (``ports''), they let `make' figure out the
> dependencies (novel!) and drag in the necessary prerequisites and they
> usually build (occasionally the gratuitous overspecificity of version
> numbers bites one in the ass; it's a real problem in the BSD/Linux
> world).

Then there's gentoo ( http://www.gentoo.org/ ) -- Linux with a ports-like
system attached. I haven't the bandwidth to use it myself, but someone at
cooper I introduced it too uses it on his home network and loves it. He
VNC'd to his computer from school (shades of plan9) to show me an 'emerge
sync' in progress.

Couldn't get him hooked on plan9, though...

--Joel