Re: [9fans] How can I shift a variable other than ?

Re: [9fans] How can I shift a variable other than ?

[erik quanstrom]

On Sat Mar 10 00:17:12 EST 2007, bsdaemon@comcast.net wrote:

> On Fri, Mar 09, 2007 at 11:46:46PM -0500, erik quanstrom wrote:
> >why wouldn't it be backwardly compatable?
> >"(a b) = $*" is currently an error in rc. and i don't
> >believe i've ever seen an rc script that depends on
> >this error. 
> 
> The problem is that then new scrips wouldn't be portable.

ya. right.  is this this the same reason i should check my c code
with the johnson c compiler (being very careful to not use function
prototypes) to make sure it works everwhere?

of course my code will probablly still be broken with the c73
compiler.  hope nobody's still running that.

and now that we've thought of it, we can't fix any rc bugs.  since
scripts that rely on fixed bugs won't run everwhere.

c'mon.  what kind of dusty-deck thinking is this?

> but in that case, you may as well just use Inferno's sh

doesn't run on plan 9.  it's written in limbo and depends on
features of inferno that are not part of plan 9.

> or port es(1)

es hasn't been maintained in a dozen years.  there is a reason
for this.

while es has some great ideas and paul haahr did a really
nice job with it.  it seems to me that let and bindings make
the shell harder to use.  es implements a lot of functonality
in es (and still breaks 10kloc of c).  here's the definition of cd.

	es% whatis cd
	@ dir{if {~ <={%count $dir} 1} {$&cd $dir} {~ <={%count $dir} 0} {%seq {if {%not {~ <={%count $home} 1}} {throw error cd <={if {~ <={%count $home} 0} {result 'cd: no home directory'} {result 'cd: home directory must be one word'}}}} {$&cd $home}} {throw error cd 'usage: cd [directory]'}}


- erik

Re: [9fans] How can I shift a variable other than ?

[Kris Maglione]

[-- Attachment #1: Type: text/plain, Size: 1422 bytes --]

On Sat, Mar 10, 2007 at 07:32:30AM -0500, erik quanstrom wrote:
>> The problem is that then new scrips wouldn't be portable.
>ya. right.  is this this the same reason i should check my c code
>with the johnson c compiler (being very careful to not use function
>prototypes) to make sure it works everwhere?

I never suggest such things.

>and now that we've thought of it, we can't fix any rc bugs.  since
>scripts that rely on fixed bugs won't run everwhere.
>
>c'mon.  what kind of dusty-deck thinking is this?

The difference here is that there has always been one rc (ignoring the 
UNIX version, which is gratuitously incompatible), unlike the bourne 
shells. One of the great things about it has always been that you could 
write rc scripts and know that they'd work on rc wherever they were run. 
If you start extending the spec, then things start to suck, truly.

>doesn't run on plan 9.  it's written in limbo and depends on
>features of inferno that are not part of plan 9.

It runs on Inferno, which runs on Plan 9. You can script for Plan 9 in 
Inferno's sh. You can even script for UNIX in it. I've done both. It 
works. It's not even ugly.

At any rate, someone might write a new shell which is neither es nor 
Inferno's shell, which is designed to run on Plan 9. I'd still rather 
just use Inferno's shell, myself.

-- 
Kris Maglione

A fool and his money soon go partying.

[-- Attachment #2: Type: application/pgp-signature, Size: 194 bytes --]

Re: [9fans] How can I shift a variable other than ?

[Dan Cross]

On 3/10/07, Kris Maglione <bsdaemon@comcast.net> wrote:
> The difference here is that there has always been one rc (ignoring the
> UNIX version, which is gratuitously incompatible), unlike the bourne
> shells. One of the great things about it has always been that you could
> write rc scripts and know that they'd work on rc wherever they were run.
> If you start extending the spec, then things start to suck, truly.

Nonsense.  You think rc has never changed before?  There have been
plenty of non-backwards compatible changes in Plan 9.

Plan 9 is a research system; it should be unfettered by the demands of
backwards compatibility with itself (within reason).  Taking your
argument to its logical conclusion, we should all be using VT220's (or
better yet, DECwriter III's) on a VAX running 7th edition Unix.  But
I'd rather incorporate good ideas for change than remain stuck in the
past.  In this environment, I think innovation is worth more than some
minimal amount o backwards compatibility.  Besides, one doesn't even
know how used this feature would be; it may come to pass that the
vast, vast majority of sites are totally unaffected.

Besides, there are sufficiently few sites running Plan 9 that, when
confronted with script breakage, it is not an unreasonable answer to
say, ``upgrade your shell.''

There are times when forward progress demands a break from established
convention.  Sometimes, this buys you nothing, but that does not mean
that every proposed change is bad.  Certainly, without deciding to
break backwards compatibility with Unix, Plan 9 would never have come
about.

> >doesn't run on plan 9.  it's written in limbo and depends on
> >features of inferno that are not part of plan 9.
>
> It runs on Inferno, which runs on Plan 9. You can script for Plan 9 in
> Inferno's sh. You can even script for UNIX in it. I've done both. It
> works. It's not even ugly.

Have to incorporate all of the Inferno machinery into a system just to
run a shell script is way too much to ask of most people.  Certainly,
more than asking them to upgrade their shell interpreter from sources.

> At any rate, someone might write a new shell which is neither es nor
> Inferno's shell, which is designed to run on Plan 9. I'd still rather
> just use Inferno's shell, myself.

Then by all means, port over the Inferno shell.

        - Dan C.

Re: [9fans] How can I shift a variable other than ?

[Kris Maglione]

[-- Attachment #1: Type: text/plain, Size: 1716 bytes --]

On Sat, Mar 10, 2007 at 04:20:22PM -0500, Dan Cross wrote:
>Nonsense.  You think rc has never changed before?  There have been
>plenty of non-backwards compatible changes in Plan 9.

You miss the point entirely. I agree that things shouldn't stay the same 
simply for the sake of compatibility. Plan 9 threw away a bunch of UNIX 
crud in the begining, and was not set in stone from the begining. I have 
no desire to add teletypes and ioctls to be compatible with UNIX.

The point is that rc(1) has been rc since the begining. It's rc 
everywhere. If you write an rc script, you can expect it work wherever 
you send it. The one infuriating exception which I've run across is the 
UNIX port, which I have to worry about being in peoples' PATHs on UNIX, 
rather than the Plan 9 version.

If there are to be changes, there's no reason to make those 
changes to rc. It would be best to create a new shell that deals with 
the shortcomings of rc, rather than adding features and cruft. That is 
how UNIX got to be so cruddy to begin with. The reason that sh(1) and 
Plan 9 are so nice is because they reevaluated and threw away most of 
what came before them.

>>It runs on Inferno, which runs on Plan 9. You can script for Plan 9 in
>>Inferno's sh. You can even script for UNIX in it. I've done both. It
>>works. It's not even ugly.

Then perhaps, as has been said for ages, it is best to make the 
integration between Inferno and Plan 9 more seemless, and the 
installation easier. It would, indeed, be nice to include Inferno in 
Plan 9 by default.

-- 
Kris Maglione

The one ingredient you made a special trip to the store
to get will be the one thing your guest is allergic to.

[-- Attachment #2: Type: application/pgp-signature, Size: 194 bytes --]

Re: [9fans] How can I shift a variable other than ?

[Dan Cross]

On 3/10/07, Kris Maglione <bsdaemon@comcast.net> wrote:
> On Sat, Mar 10, 2007 at 04:20:22PM -0500, Dan Cross wrote:
> >Nonsense.  You think rc has never changed before?  There have been
> >plenty of non-backwards compatible changes in Plan 9.
>
> You miss the point entirely. I agree that things shouldn't stay the same
> simply for the sake of compatibility. Plan 9 threw away a bunch of UNIX
> crud in the begining, and was not set in stone from the begining. I have
> no desire to add teletypes and ioctls to be compatible with UNIX.
>
> The point is that rc(1) has been rc since the begining. It's rc
> everywhere. If you write an rc script, you can expect it work wherever
> you send it. The one infuriating exception which I've run across is the
> UNIX port, which I have to worry about being in peoples' PATHs on UNIX,
> rather than the Plan 9 version.
>
> If there are to be changes, there's no reason to make those
> changes to rc. It would be best to create a new shell that deals with
> the shortcomings of rc, rather than adding features and cruft. That is
> how UNIX got to be so cruddy to begin with. The reason that sh(1) and
> Plan 9 are so nice is because they reevaluated and threw away most of
> what came before them.

No, I didn't miss the point, which, if carried to its logical
conclusion would imply that we should just write a new operating
system whenever we want to add something new.  There's nothing set in
stone about any system; the problems you describe come when people
start believing that there is.  The proliferation of shells under Unix
could have been avoided by carefully re-evaluating the existing offers
and making appropriate changes.  Instead, we ended up with Shell Soup.

Rc is just a program; yes, a pretty good one, but it is not a
religious document: it is open to interpretation, re-evaluation and
change.

Put another way, one of the reasons Plan 9 has remained so nice over
the years is because the folks primary responsible for it have been
open to making reasonable changes where appropriate.

> >>It runs on Inferno, which runs on Plan 9. You can script for Plan 9 in
> >>Inferno's sh. You can even script for UNIX in it. I've done both. It
> >>works. It's not even ugly.
>
> Then perhaps, as has been said for ages, it is best to make the
> integration between Inferno and Plan 9 more seemless, and the
> installation easier. It would, indeed, be nice to include Inferno in
> Plan 9 by default.

You're replying to yourself there.

        - Dan C.

Re: [9fans] How can I shift a variable other than ?

[matt]

 > You miss the point entirely.

This might be petty but, no Kris, you missed the point.

If you want further proof, see rc(1) BUGS

Your broken rc will not interpret scripts intended for the One True rc

When you've got ghosts in /bin, who you gonna call ?

Re: [9fans] How can I shift a variable other than ?

[Kris Maglione]

[-- Attachment #1: Type: text/plain, Size: 877 bytes --]

On Sun, Mar 11, 2007 at 12:12:57PM +0000, matt wrote:
> > You miss the point entirely.
>
>This might be petty but, no Kris, you missed the point.
>
>If you want further proof, see rc(1) BUGS

I never said that rc(1) was perfect. I'm all for rewriting it, and 
fixing bugs. I'm not for breaking its syntax, though. It makes far more 
sense to leave rc as rc(1) and to write a new shell which deals with the 
shortcomings of rc. This is not comprable to writing a new os, and it 
gives us far more freedom. If the Inferno people had decided to 'update' 
mash instead of writing sh(1), it would have been grossly unfortunate, 
because Inferno's shell is probably the best shell I've come across.

And if rc is to have such a massive revamping, there's no good reason to 
keep calling it rc.

-- 
Kris Maglione

The faster the plane,
the narrower the seats.

[-- Attachment #2: Type: application/pgp-signature, Size: 194 bytes --]

Re: [9fans] How can I shift a variable other than ?

[erik quanstrom]

On Sun Mar 11 15:28:11 EST 2007, bsdaemon@comcast.net wrote:
> I never said that rc(1) was perfect. I'm all for rewriting it, and 
> fixing bugs. I'm not for breaking its syntax, though. It makes far more 
> sense to leave rc as rc(1) and to write a new shell which deals with the 
> shortcomings of rc. 

rc is not the unchanging point at the center of a circle you think it is.  
<> redirection and $" are not in the original rc paper.

and more to the point, there have been no ideas yet that would
break older rc scripts nor change the grammer or lexemes.  even list
assignment would be a matter of handling a case that currently gives
an error.  the grammer supports it.

the things i've added experimentally are

1.  support for history.
2.  a break statement.
3.  a fix to make wait reliable as per the december discussion.
4.  subscript ranges -- $fu(n-m)

of these, only break even goes so far as to add a keyword.

- erik

Re: [9fans] How can I shift a variable other than ?

[Martin Neubauer]

* erik quanstrom (quanstro@coraid.com) wrote:
> and more to the point, there have been no ideas yet that would
> break older rc scripts nor change the grammer or lexemes.  even list
> assignment would be a matter of handling a case that currently gives
> an error.  the grammer supports it.

Specifically, there are two arguments supporting the change. First, it isn't
really a new feature -- it just makes one already present more general (with
a striking resemblance to the for loop.) Second, it doesn't break scripts in a
harmful way (old scripts still run, new scripts run in an old rc abort -- if
a correctly written script runs successfully it does what was intended in
either case.)

	Martin

Re: [9fans] How can I shift a variable other than ?

[Kris Maglione]

[-- Attachment #1: Type: text/plain, Size: 1545 bytes --]

On Mon, Mar 12, 2007 at 12:14:08AM +0100, Martin Neubauer wrote:
>Specifically, there are two arguments supporting the change. First, it isn't
>really a new feature -- it just makes one already present more general (with
>a striking resemblance to the for loop.) Second, it doesn't break scripts in a
>harmful way (old scripts still run, new scripts run in an old rc abort -- if
>a correctly written script runs successfully it does what was intended in
>either case.)

This is entirely beside the point. Breaking old scripts is the least of 
the issues. The point is that there's little value in altering rc, 
compared to writing a new shell. Compare mash and sh on Inferno and 
you'll see what I mean, as I've said. If mash had simply been extended, 
there would be either a slightly more or sligtly less crufty mash, 
instead of sh.

If rc is to be updated, it should simply be replaced with something 
better. If Plan 9 is a research OS, as has been suggested in this 
thread, then why are we to add features to an old shell, rather than 
rethinking it? The former has nothing to do with research. I'm too tired 
to argue saliently, but the point is that if rc is lacking, we've been 
shown better ways already, and they should be instrumented or improved 
upon properly, not simply hacked onto old cruft.

Again, this is exactly how UNIX ended up how it is today, and why Plan 9 
is not UNIX.

-- 
Kris Maglione

An easily-understood, workable falsehood is more useful
than a complex, incomprehensible truth.

[-- Attachment #2: Type: application/pgp-signature, Size: 194 bytes --]

Re: [9fans] How can I shift a variable other than ?

[erik quanstrom]

On Mon Mar 12 03:51:10 EST 2007, bsdaemon@comcast.net wrote:

> On Mon, Mar 12, 2007 at 12:14:08AM +0100, Martin Neubauer wrote:
> This is entirely beside the point. Breaking old scripts is the least of 
> the issues. The point is that there's little value in altering rc, 
> compared to writing a new shell. Compare mash and sh on Inferno and 
> you'll see what I mean, as I've said. If mash had simply been extended, 
> there would be either a slightly more or sligtly less crufty mash, 
> instead of sh.

you've said this multiple times, and we still disagree.

> 
> If rc is to be updated, it should simply be replaced with something 
> better. If Plan 9 is a research OS, as has been suggested in this 
> thread, then why are we to add features to an old shell, rather than 
> rethinking it? The former has nothing to do with research. I'm too tired 
> to argue saliently, but the point is that if rc is lacking, we've been 
> shown better ways already, and they should be instrumented or improved 
> upon properly, not simply hacked onto old cruft.

it sounds like you're the right man for the job.  why don't you get
started?  i'll be happy to test it when it has a manual page and the
silly bugs knocked out.

- erik

Re: [9fans] How can I shift a variable other than ?

[Anthony Sorace]

what's being proposed here for addition to rc is totally different
from the mash/inferno-sh case.

in the inferno case, we had two different authors with some
substantial differences in ideas for constructing shells, in terms of
syntax, semantics, and features. you're correct that trying to add
rog's (right?) ideas into mash would've been a disaster, and i think
everyone here would agree that we shouldn't try to cram that sort of
difference in the fundamentals of building shells into rc.

but that simply isn't what's being proposed. the feature to be added
is entirely in line with rc's existing grammar and design principles.
it's a much more incremental type of growth, and is an equally valid
result of research.

i understand that your position does leave room for moving off vt220s,
but it does seem to imply that we shouldn't add new kernel devices
without scrapping plan9 and starting over.

i also find it bizarre that you can call rc "old cruft"...

anthony

Re: [9fans] How can I shift a variable other than ?

[Dan Cross]

On 3/11/07, Kris Maglione <bsdaemon@comcast.net> wrote:
> [...]
>
> And if rc is to have such a massive revamping, there's no good reason to
> keep calling it rc.

I think the point you keep missing is that this isn't such a massive
revamping.  It's a simple change that doesn't break any existing
scripts; the worst thing that could happen is that new scripts don't
run on the old shell, in which case one can update the shell.

This really isn't that big of a deal.  You're making a mountain out of
a molehill.  As for your point about writing a new shell.  Well, go do
it, then.

        - Dan C.

Re: [9fans] How can I shift a variable other than ?

[Russ Cox]

Everyone has said their piece,
and opinions have stopped changing.
Let's move on.

Russ

Re: [9fans] How can I shift a variable other than ?

[Dan Cross]

Was that a Haiku?

On 3/12/07, Russ Cox <rsc@swtch.com> wrote:
> Everyone has said their piece,
> and opinions have stopped changing.
> Let's move on.
>
> Russ
>

Re: [9fans] How can I shift a variable other than ?

[Kris Maglione]

[-- Attachment #1: Type: text/plain, Size: 262 bytes --]

On Mon, Mar 12, 2007 at 01:05:24PM -0400, Dan Cross wrote:
>Was that a Haiku?

Yes. At any rate, I'm tired of trolling. I'll stop until I start
seeing things in /n/sources/patch.

-- 
Kris Maglione

Whatever it is, somebody will have had it for lunch.

[-- Attachment #2: Type: application/pgp-signature, Size: 194 bytes --]

Re: [9fans] How can I shift a variable other than ?

[rog]

> If the Inferno people had decided to 'update' mash instead of writing sh(1)

for the record, i believe that i wrote the inferno shell before mash was written
(in fact, it was one of the first limbo programs i wrote, before i came to VN).
certain of its features were only made possible by later inferno changes
however (e.g. the shell module loading structure was enabled by the addition of load self).
i have to say i wouldn't mind seeing a version for plan 9 - the lack of support for quoting
and unquoting lists i find particularly awkward in rc in these days of
space-containing filenames. i still wish plan 9 hadn't relaxed the whitespace ban in names!

somebody (anothy?) made a comment ages ago about how it was "suprisingly slow". i'm not
surprised - i made no attempt at all to be clever about the interpretation strategy
(unlike rc, which builds an intermediate language), and lists are reversed willy nilly.
variable access could probably be speeded up a fair bit.
unfortunately it's probably not possible to improve things without breaking
binary compatibility.

re: the shift thing: changing a feature is always a trade-off between backward
compatibility and feature benefits. in this case, the feature is still available
(if a little more awkwardly), in the original, so i think backward compatibility
considerations win. the I/O feature, <>, on the other hand, provides useful (sometimes
essential) functionality not previously available, so the balance swings
the other way.

Re: [9fans] How can I shift a variable other than ?

[erik quanstrom]

strictly speaking, <> could be implemented with no help
from rc (and more generally) as in

	boxredir fdin fdout cmd

okay, that's a bit silly, but the point between essential and
non-essential is in the eye of the beholder.

odd digression:

in the late 60s the fad was to rewrite bigger instruction sets in
terms of smaller ones (the metric was # of instructions).  somebody
had a 16 instruction set that someone else reimplemented in 8
instructions.  the apollo guidance computer had 12 instructions and
a few magic memory locations.  it was finally determinted you could get by on
one instruction --- some variation on predecrement and branch
if negative.  "look ma, no instructions."  unfortunately, this machine
was somewhat difficult to program

- erik

Re: [9fans] minimal-instruction-sets (was: How can I shift a variable other than ?)

[andrey mirtchovski]

> in the late 60s the fad was to rewrite bigger instruction sets in
> terms of smaller ones (the metric was # of instructions).  somebody
> had a 16 instruction set that someone else reimplemented in 8
> instructions.  the apollo guidance computer had 12 instructions and
> a few magic memory locations.  it was finally determinted you could  
> get by on
> one instruction --- some variation on predecrement and branch
> if negative.  "look ma, no instructions."  unfortunately, this machine
> was somewhat difficult to program
>
> - erik

i think you're referencing the One Instruction Set Computer (OISC).

http://en.wikipedia.org/wiki/OISC

Re: [9fans] How can I shift a variable other than ?

[C H Forsyth]

>> If the Inferno people had decided to 'update' mash instead of writing sh(1)
>for the record, i believe that i wrote the inferno shell before mash was written

nothing to do with 9fans, but since the matter has surfaced here:
the only real trouble with mash for Vita Nuova at the time was that the group
from which we acquired Inferno in its fire sale was not really on speaking terms
with the group that by then contained its author (brucee),
and rather than creating an incompatible variant of mash,
we had roger continue work on his, since we needed something more than
the original Inferno sh (or felt we did).  despite brucee's own efforts we never did
get even as far as a talk about talks with the right people in the other groups involved.

anyway, as to the archaic nature of shells.
	>i also find it bizarre that you can call rc "old cruft"...

i supposed that was a reference to the fact that the style of these shells hasn't
changed all that much since 1977.

Re: [9fans] How can I shift a variable other than ?

[Anthony Sorace]

On 3/13/07, C H Forsyth <forsyth@vitanuova.com> wrote:
> anyway, as to the archaic nature of shells.
>         >i also find it bizarre that you can call rc "old cruft"...
>
> i supposed that was a reference to the fact that the style of these shells hasn't
> changed all that much since 1977.

interestingly, the most different shell i've seen is Windows
PowerShell (formerly MSH, aka Monad). not to say i'm particularly a
fan, but the idea of an OO CLI is interesting.

Re: [9fans] How can I shift a variable other than ?

[Paweł Lasek]

On 3/13/07, Anthony Sorace <anothy@gmail.com> wrote:
> On 3/13/07, C H Forsyth <forsyth@vitanuova.com> wrote:
> > anyway, as to the archaic nature of shells.
> >         >i also find it bizarre that you can call rc "old cruft"...
> >
> > i supposed that was a reference to the fact that the style of these shells hasn't
> > changed all that much since 1977.
>
> interestingly, the most different shell i've seen is Windows
> PowerShell (formerly MSH, aka Monad). not to say i'm particularly a
> fan, but the idea of an OO CLI is interesting.

AFAIK isn't it the only "normal" official way of using the whole NT
namespace? I always find the fact that NT uses Unix-style files
(complete with ioctl :D) for device access but that part of namespace
is usually hidden deep inside...

Although what they are doing for Linux-based PalmOS replacement can
rival it... (It includes completely different system - it has unix
kernel but afaik uses many concepts alien to unix as base)

-- 
Paul Lasek

Re: [9fans] How can I shift a variable other than ?

[Anthony Sorace]

On 3/13/07, rog@vitanuova.com <rog@vitanuova.com> wrote:
> somebody (anothy?) made a comment ages ago about how it was "suprisingly slow".

that was me. don't misunderstand: i quite like the shell, and that
wasn't intended to imply that it's not suitable for a great number of
things (including, of course, interactive use and one-offs). i suppose
i was just disappointed that we couldn't yet really use shell scripts
as extensively as they are in plan 9.

> unfortunately it's probably not possible to improve things without breaking
> binary compatibility.

binary compatibility with what?