Re: [9fans] Getting started in Plan9 - help

Re: [9fans] Getting started in Plan9 - help

[Russ Cox]

First of all, I was kidding about readline.

> * If dynamic linked libraries, you have taken the first step
>   on the slippery slope to the land of gratuitious complexity,
>   where even a simple "hello, world" C program can execute
>   100000 instructions before it actually prints anything.
>   (No, I'm not kidding.)  For some reason, people seem to think
>   adding features in dynamic libraries is "free".

If I'm running rio on Plan 9, draw a new window, and run
hello world, how many instructions execute before I see
"hello world" on my screen?  Lots and lots, just not many
in the hello world process.  The kernel has to redirect
the write into rio, rio does some lookups to figure out
which window gets the output, libdraw has to walk around
in the font caches, possibly loading the missing letters,
(which is more kernel calls, context switches to kfs,
and execution of the various code in devsd), finally
preparing a "draw string" request for /dev/draw.  Then the
kernel is back.  For simplicity let's assume that the window
is in front of all the others, so memlayer isn't involved.  Now we
clear the background, either with memset or by making
calls to an accelerated vga routine with some busy waits
(still faster than memset, though).  Then we use
per-character calls to memimagedraw to blacken the appropriate
pixels.  And finally we see "hello world" (assuming that we
have direct access to video memory; if not, we've been drawing
in a soft screen and still need to flush the changed portion
out to video memory).

You're still right, of course; dynamic libraries are a
slippery slope for lots of reasons.  My point is that the
overhead of dynamic libraries is not necessarily worse than
the overhead of other system services like the kernel and
the various Plan 9 file servers, just in a different place.
It takes far more than 100,000 instructions to get "hello world"
onto my screen either way.  How you count depends on what
you care to include.

Russ

Re: [9fans] Getting started in Plan9 - help

[Mike Haertel]

>>   where even a simple "hello, world" C program can execute
>>   100000 instructions before it actually prints anything.
>
>If I'm running rio on Plan 9, draw a new window, and run
>hello world, how many instructions execute before I see
>"hello world" on my screen?  Lots and lots, just not many
>in the hello world process.  The kernel has to redirect
>the write into rio, rio does some lookups to figure out
>which window gets the output, [...]
>
>It takes far more than 100,000 instructions to get "hello world"
>onto my screen either way.  How you count depends on what
>you care to include.

Under an operating system that will remain nameless, the 100K figure
came from using a debugger to count only instructions in the process
context of the "hello, world" program itself.  It didn't count
kernel mode instructions executed, or window system operations in
other processes, or any of the plethora of things you mentioned.
Had I counted those it of course would have been much higher.

Writing to a file or pipe is much cheaper and there the 100K
instructions start to stick out like a sore thumb.

The reason I bitch and whine about the 100K instructions (or lots
more than 100K in some operating systems) is because process startup
overhead is one of the main limiting factors to the performance of
shell scripts, mkfiles, and the like.  That in turn makes people
give up on the "tools" approach too soon and too often.

Re: [9fans] Getting started in Plan9 - help

[Douglas A. Gwyn]

Boyd Roberts wrote:
> Depends on your domain.  In mathematics a product could be a dot
> [rc's ^] or a cross product [trickier].

There are all sorts of "products" in mathematics,
but the so-called "cross product" is one of the least qualified
to go by that name.

Re: [9fans] Getting started in Plan9 - help

[Boyd Roberts]

> In formal languages theory, "product" is usually concatenation,
> "exponentiation" is repeated concatenation.

Depends on your domain.  In mathematics a product could be a dot
[rc's ^] or a cross product [trickier].

Re: [9fans] Getting started in Plan9 - help

[Dan Cross]

> In formal languages theory, "product" is usually concatenation,
> "exponentiation" is repeated concatenation.  (For example, one says
> that a DFA can scan A^n but not A^n B^n; one says that a context free
> grammar can parse A^n B^n but not A^n B^n C^n.)

Whoops, I should have realized that's what you were refering to.  My bad.

	- Dan C.

Re: [9fans] Getting started in Plan9 - help

[Ralph Corderoy]

Hi Matt,

> The downside is that I have to pass the username & password on the
> command line too so it starts to get a bit unwieldy :

You could put them in ~/.my.cnf.

    [client]
    user=xxx
    password=yyy


Ralph.

Re: [9fans] Getting started in Plan9 - help

[Douglas A. Gwyn]

"Thomas Bushnell, BSG" wrote:
> The ASCII BS charactec is specifically to help the "backup and
> overstrike" operation.

Actually it's just the backspace part of that composite operation,
which is what we said.

> That is, DEL is *overstruck* on a character to delete it.  Note the
> nice way this works on paper tape: DEL is all-ones, so that means a
> paper tape reader should ignore all-ones.  It's also supposed to
> ignore NUL (all-zeros); some hardware could not be relied upon to deal
> so accurately with spacing, and false reads of all-zeros would happen.

That's a bit revisionist; no paper tape punch that I know of
ever automatically reversed tape motion upon receipt of a BS
character.  The overstrike-DEL-to-delete was done *manually*
when preparing tapes off line.  The ignore-NUL mode was normally
used only for the *initial* mode of each tape ("file"), i.e.
to skip leader, with embedded NUL characters often used in binary
encodings.

Another cute thing:  The ASR-33 didn't immediately stop the reader
when an XOFF was read, and it didn't even stop reliably after the
*next* character had already passed, unless that character was a
DEL.  (Apparently when all 8 sensing fingers activated there was a
big enough dump of mechanical energy to ensure reliable stopping.)
Therefore, Honeywell's source tape convention was LF, text, CR,
XOFF, DEL (usually called RUBOUT rather than DEL).  The funny
split of the LF and CR actions allowed the assembler to input a
line from the reader with echo-print on the paper, then starting
from the new (beginning-of-line) position it would print the
object code, at some point sending a TAPE-ON (DC2) to enable the
punch unit and then the binary record, which used all nonprinting,
nonmotion characters (including BEL, ding ding) then TAPE-OFF (DC4).
At which point XON (DC1) would be sent to enable the reader which
would input the next text line (starting with that echoed LF to
clear the just-completed line on the paper).  A nifty way to use
the ASR-33 for a complete one-pass assembly process with object
code and full output listing.  You can investigate this using
one of the Honeywell Series-16 (+ ASR-33) simulators (Sapere's or
Supnick's) and the DAP assembler (available at Sapere's site).

Re: [9fans] Getting started in Plan9 - help

[Thomas Bushnell, BSG]

cross@math.psu.edu (Dan Cross) writes:

> > To delete n characters, obviously, you sound BS^n DEL^n, and certainly
> > not (BS DEL)^n.
>
> I think you mean BS * n followed by DEL * n.  You want repitition, not
> exponentiation (which would be meaningless, I think).  The operation in
> this context is non-distributive, though.

In formal languages theory, "product" is usually concatenation,
"exponentiation" is repeated concatenation.  (For example, one says
that a DFA can scan A^n but not A^n B^n; one says that a context free
grammar can parse A^n B^n but not A^n B^n C^n.)

Your notation would have "addition" to be concatenation, and "product"
to be repeated concatenation.  That convention would work as well, but
since addition is conventionally always commutative, and product is
not, it's a more confusing convention.  (As you note, it suggests a
distributive property which of course does not exist.)

In other words, we agree about substance but are just using different
notations.

> > There may have been additional meanings to DEL, but the only one I
> > know of is as part of a BS DEL sequenc.e.
>
> Is this one of those, ``well, people are still posting about this
> subject, therefore I feel no need to worry about its topicality...''
> things?

Of course.  :)  Plus I think it's fun.

> Surely there are more important things to worry about, like getting
> the Atari 2600 simulator working....

Well, or getting Plan 9 to boot on my laptop.

Re: [9fans] Getting started in Plan9 - help

[Dan Cross]

> To delete n characters, obviously, you sound BS^n DEL^n, and certainly
> not (BS DEL)^n.

I think you mean BS * n followed by DEL * n.  You want repitition, not
exponentiation (which would be meaningless, I think).  The operation in
this context is non-distributive, though.

> There may have been additional meanings to DEL, but the only one I
> know of is as part of a BS DEL sequenc.e.

Is this one of those, ``well, people are still posting about this
subject, therefore I feel no need to worry about its topicality...''
things?

Surely there are more important things to worry about, like getting
the Atari 2600 simulator working....

	- Dan C.

Re: [9fans] Getting started in Plan9 - help

[Thomas Bushnell, BSG]

arnold@dstc.monash.edu.au (David Arnold) writes:

> -->"Thomas" == Thomas Bushnell, BSG <tb+usenet@becket.net> writes:
>
>   Thomas> ASCII has no erase character (just as it has no newline
>   Thomas> character), and there is no "one true character" for erase,
>   Thomas> any more than there is one true C brace indenting style or
>   Thomas> one true endianness.
>
> just to expose my ignorance, what was the intended meaning of DEL (127) ?

As I mentioned before, DEL and NUL were the two ignored characters.

To tell the receiver to erase the last received character, you would
send a BS DEL sequence; that is, you'd overstrike the last character
with a DEL character.  On paper tape, that did exactly the right thing
very cleverly without any special handling; other media were supposed
to DTRT.

To delete n characters, obviously, you sound BS^n DEL^n, and certainly
not (BS DEL)^n.

There may have been additional meanings to DEL, but the only one I
know of is as part of a BS DEL sequenc.e.

Thomas

Re: [9fans] Getting started in Plan9 - help

[Thomas Bushnell, BSG]

rsc@plan9.bell-labs.com (Russ Cox) writes:

> > I have way more keys that ASCII has codes for.  How exactly should
> > that work?  X has a good way (even Lispms had a good way).
>
> Arbitrarily configurable does not imply good.
> It does usually imply more pain for users.

Arbitrarily configurable is usually wrong.  I don't mean that
everything about X is good.

Just that it has a notion of "keycode" and that's what a key sends,
rather than trying to force ASCII to do the job.

Thomas

Re: [9fans] Getting started in Plan9 - help

[Thomas Bushnell, BSG]

boyd@strakt.com (Boyd Roberts) writes:

> "Thomas Bushnell, BSG" wrote:
> > One misunderstanding is that "ASCII doesn't represent accents", which
> > is totally wrong; it was deliberately constructed to allow for accents
> > for all/many languages that use the latin alphabet.  The idea was (as
> > I mentioned previously) to use overstriking.
> >
> > Now it turns out that the idea was not as good ...
>
> s/not as good/ghastly/

Well, there are lots of multibyte fans out there still.   But hey, it
was really the very first guess; I think they should get *some*
credit. :)

Re: [9fans] Getting started in Plan9 - help

[Russ Cox]

> I have way more keys that ASCII has codes for.  How exactly should
> that work?  X has a good way (even Lispms had a good way).

Arbitrarily configurable does not imply good.
It does usually imply more pain for users.

Russ

Re: [9fans] Getting started in Plan9 - help

[Boyd Roberts]

"Thomas Bushnell, BSG" wrote:
> One misunderstanding is that "ASCII doesn't represent accents", which
> is totally wrong; it was deliberately constructed to allow for accents
> for all/many languages that use the latin alphabet.  The idea was (as
> I mentioned previously) to use overstriking.
>
> Now it turns out that the idea was not as good ...

s/not as good/ghastly/

Re: [9fans] Getting started in Plan9 - help

[David Arnold]

-->"Thomas" == Thomas Bushnell, BSG <tb+usenet@becket.net> writes:

  Thomas> ASCII has no erase character (just as it has no newline
  Thomas> character), and there is no "one true character" for erase,
  Thomas> any more than there is one true C brace indenting style or
  Thomas> one true endianness.

just to expose my ignorance, what was the intended meaning of DEL (127) ?




d

Re: [9fans] Getting started in Plan9 - help

[Thomas Bushnell, BSG]

boyd@strakt.com (Boyd Roberts) writes:

> "Thomas Bushnell, BSG" wrote:
> > The ASCII BS charactec is specifically to help the "backup and
> > overstrike" operation.
>
> Some unix tty drivers would do this.  They'd overstrike with a space
> and then a BS :)

Some overstrikes are easier than others.

There are two misunderstandings about ASCII that piss me off, that's
all.

One misunderstanding is that "ASCII doesn't represent accents", which
is totally wrong; it was deliberately constructed to allow for accents
for all/many languages that use the latin alphabet.  The idea was (as
I mentioned previously) to use overstriking.

Now it turns out that the idea was not as good as just separate codes
for the accents (though that has its problems too), but it's not like
there was some American chauvinism that didn't bother to think about
other languages.

The other thing that pisses me off is people complaining that this or
that character is "the erase character" in ASCII, and complaining that
some system is using a different character.  ASCII has no erase
character (just as it has no newline character), and there is no "one
true character" for erase, any more than there is one true C brace
indenting style or one true endianness.

Thomas

Re: [9fans] Getting started in Plan9 - help

[Thomas Bushnell, BSG]

rsc@plan9.bell-labs.com (Russ Cox) writes:

> Further, you completely missed the point.
> dhog claimed that typing a backspace key should
> send a backspace character, just like typing an
> 'a' sends an 'a' character; the ascii
> semantics of backspace are not relevant
> to that particular claim.

I have way more keys that ASCII has codes for.  How exactly should
that work?  X has a good way (even Lispms had a good way).

Re: [9fans] Getting started in Plan9 - help

[Thomas Bushnell, BSG]

rminnich@lanl.gov (Ronald G Minnich) writes:

> yeah but I don't care. I want the keyboard to do what I want, not ride
> some standard defined when we all used KSRs for terminals. I dismantled
> and threw my Teletypes out 25 years ago (although I saved some pictures --
> heroic hardware, that stuff was).

Sure, but to get the keyboard to "do what you want" doesn't say
anything about what internal coding the key sends.

I want the "big key with the arrow" to delete the previous character.
I couldn't care less whether that is internally coded with 010 or 0177.

Thomas

Re: [9fans] Getting started in Plan9 - help

[Lyndon Nerenberg]

>>>>> "Douglas" == Douglas A Gwyn <DAGwyn@null.net> writes:

    Douglas> If you're going
    Douglas> to pay for Unity you might consider licensing the whole
    Douglas> Unix System ToolChest; I think there was a good price for
    Douglas> this last time I checked.  There are some gems in there,
    Douglas> including algorithm animation and geo-mapping utilities.

Pity, they seem to have dropped the System V rogue (and HDB) :-(

DWB is there, but can you license the Writers Workbench from anywhere
these days?

--lyndon

Re: [9fans] Getting started in Plan9 - help

[Boyd Roberts]

"Thomas Bushnell, BSG" wrote:
> The ASCII BS charactec is specifically to help the "backup and
> overstrike" operation.

Some unix tty drivers would do this.  They'd overstrike with a space
and then a BS :)

Re: [9fans] Getting started in Plan9 - help

[Russ Cox]

> The ASCII standard fortunately solves this one, and you're both wrong.

I'll keep that in mind the next time
I port Plan 9 to a machine that runs
off paper tape.  Thanks.

> > the default should be that backspace is backspace.

Further, you completely missed the point.
dhog claimed that typing a backspace key should
send a backspace character, just like typing an
'a' sends an 'a' character; the ascii
semantics of backspace are not relevant
to that particular claim.

Russ

Re: [9fans] Getting started in Plan9 - help

[Scott Schwartz]

> The ASCII standard

I thought Plan 9 had given up on that? :-)

Re: [9fans] Getting started in Plan9 - help

[Ronald G Minnich]

On Wed, 23 Jan 2002, Thomas Bushnell, BSG wrote:

> The ASCII standard fortunately solves this one, and you're both wrong.

yeah but I don't care. I want the keyboard to do what I want, not ride
some standard defined when we all used KSRs for terminals. I dismantled
and threw my Teletypes out 25 years ago (although I saved some pictures --
heroic hardware, that stuff was).

ron

Re: [9fans] Getting started in Plan9 - help

[Thomas Bushnell, BSG]

dhog@plan9.bell-labs.com (David Gordon Hogan) writes:

> Of course, xterm is configurable through some resource, but
> the default should be that backspace is backspace.

You mean, the default is that "backspace erases a character backwards"
or "backspace sets up an overstrike"?  The latter is what ASCII means
by backspace.

Or maybe you mean that the default is that "backspace should send #".

Re: [9fans] Getting started in Plan9 - help

[Thomas Bushnell, BSG]

"Douglas A. Gwyn" <DAGwyn@null.net> writes:

> David Gordon Hogan wrote:
> > What _really_ annoys me is when I hit the BACKSPACE key on
> > the keyboard, while running some terminal emulator, and it
> > sends DELETE.
> 
> DEC actually started making terminals like that!
> 
> > the default should be that backspace is backspace.
> 
> I agree..

The ASCII standard fortunately solves this one, and you're both wrong.

The ASCII BS charactec is specifically to help the "backup and
overstrike" operation.  For example, à is supposed to be rendered as 

  a BS `

The ASCII sequence for "ignore last character transmitted" is:

  BS DEL

That is, DEL is *overstruck* on a character to delete it.  Note the
nice way this works on paper tape: DEL is all-ones, so that means a
paper tape reader should ignore all-ones.  It's also supposed to
ignore NUL (all-zeros); some hardware could not be relied upon to deal
so accurately with spacing, and false reads of all-zeros would happen.

Thomas

Re: [9fans] Getting started in Plan9 - help

[Boyd Roberts]

David Gordon Hogan wrote:
> What _really_ annoys me is when I hit the BACKSPACE key on
> the keyboard, while running some terminal emulator, and it
> sends DELETE.  This has a tendency to make me feel homicidal.

I'm with you Captain, I can already smell the napalm burning...

I tried to fight that fight with a cordless Logitech iTouch,
X11, xterm and xmodmap and gave up in disgust at some point
last week.

The _really freaky thing_ is that when I type 8 bit chars
on my Swedish qwerty keyboard _in some windows_ I see
Cyrillic.  I can read it, but it doesn't make matters
any better.

I won't even touch KDE.  If I want a window [xterm] I use 'x':

    x [ hostname ]

Actually I'd like to build jammers for these horrible cordless
things :)

Re: [9fans] Getting started in Plan9 - help

[Douglas A. Gwyn]

skipt@real.com wrote:
> unity, ...

As I recall the source code had some problems with argument testing
(essentially, dereferencing null pointers before testing), easily
fixed once you're aware.  If you're going to pay for Unity you might
consider licensing the whole Unix System ToolChest; I think there
was a good price for this last time I checked.  There are some gems
in there, including algorithm animation and geo-mapping utilities.

Re: [9fans] Getting started in Plan9 - help

[Bakul Shah]

ozan s yigit wrote:
> it is "reactive keyboard" and i believe was a thesis work at university
> of calgary, by Darragh under Witten. i'm sure a web search would still turn
> up pointers. there is a book about it, not sure if still in print. the
> interface was interesting in trying to accomodate disabled people to
> interact with command interfaces by predictive completion.

See volume 20 issues 29..32 of comp.sources.unix (Oct '89) at
  ftp://gatekeeper.dec.com/pub/usenet/comp.sources.unix/volume20/reactivekbd
The most recent version seems to be at
  http://www.csoft.net/~dummy/robert/software/rk-1.6.2.tar.gz

Re: [9fans] Getting started in Plan9 - help

[Douglas A. Gwyn]

David Gordon Hogan wrote:
> What _really_ annoys me is when I hit the BACKSPACE key on
> the keyboard, while running some terminal emulator, and it
> sends DELETE.

DEC actually started making terminals like that!

> the default should be that backspace is backspace.

I agree..

Re: [9fans] Getting started in Plan9 - help

[skipt]

unity, a package that could be found at AT&T Toolchest long ago, kind of
did relational data management the "unix" way. My memory is not very clear
on this, but I recall it used text files, with hash index files for fast searches, and
I think it embedded the dictionary as the first (comment) line of
each file. You could use grep and awk, etc. I used it with some large data
sets, and it worked correctly and was fast. (I just checked, Lucent SSG sells it)

It is much harder to tastefully map a  table (matrix structure) to
a filesystem (a tree).

At 03:39 PM 1/21/2002 -0500, Russ Cox wrote:
>I think it might be neat to build a small number
>of little programs that parse and reemit sql queries.
>Thus you could use normal shell pipelines to do database
>queries but since you're passing _queries_ and not _output_
>through the pipes, you still reap the benefits of
>query optimization.  Just like the old /acme/edit scripts.
>
>I don't remember enough of the SQL syntax to
>reparse your query into an exact shell pipeline, but
>the idea is you could do something like
>
>	sql/grep 'Car%' pages.title | sql/join section pages.idx |
>	sql/p
>
>Sql/p would behave like the acme p, actually executing
>the query to produce output.
>
>Of course, if there isn't much data, moving it into
>the file system might be an even better solution.
>
>Russ
>
>

Re: [9fans] Getting started in Plan9 - help

[David Gordon Hogan]

> On Unix systems since around 7th Edition, one can bind any input
> character to the INTR function, and the DEL character was the
> default (somewhat confused by Berkeleyites who tried to change
> to DEC OS conventions: ^C -> INTR, DEL -> char-erase, etc.).
> ^? is just a way of representing the ASCII DEL character using
> printable notation; the actual character is still DEL (0x7F).

What _really_ annoys me is when I hit the BACKSPACE key on
the keyboard, while running some terminal emulator, and it
sends DELETE.  This has a tendency to make me feel homicidal.

Actually, xterm is doing it to me right now.  xev receives the
keystroke correctly as "keycode 64 (keysym 0xff08, BackSpace)",
then xterm says, duh, I think you want delete.  Grrrr!

Of course, xterm is configurable through some resource, but
the default should be that backspace is backspace.

Re: [9fans] Getting started in Plan9 - help

[Douglas A. Gwyn]

Boyd Roberts wrote:
> Yes, I liked it too, so much so I hacked the SysVr2 shell
> to do it and got it to do 'whatis', chucked out 'type'
> and added 'builtin'.

The BRL (now part of ARL) Bourne shell already had those,
also a $HISTORY file (called $HISTFILE to avoid conflict),
emacs-style command line editing, etc.  In fact I'm still
using it; we have a single source that can be conditionally
configured to fit any known Unix platform, which vary a lot
more than one might think.  It also supports csh-like job
control, and as a side effect uncovered a very bad,
pervasive race in the Irix kernel (which was soon fixed).

Re: [9fans] Getting started in Plan9 - help

[Matt H]

On Tue, 22 Jan 2002 16:36:15 +0100
"cej@cejchan.gli.cas.cz" <cej@cejchan.gli.cas.cz> wrote:

> >
> > Of course, if there isn't much data, moving it into
> > the file system might be an even better solution.
> >
> > Russ
>
> Could anyone show me how to do such a thing? (tables + relations --> filesys)

man join
man awk
man sort
man bind
and some custom scripting is about as far as I've got

my data is such that ls /data/content/*/*.txt works for me

I'm trying to keep my data with a depth of one although I envisage being able to move to a more tree like structure to resolve inner joins and the like.

Of course, you have to consider the performance issue of losing your indexes.

Hopefully I'll be working a bit harder on it in a couple of weeks time when my current nightmare has ended !

M

Re: [9fans] Getting started in Plan9 - help

[cej]

>
> Of course, if there isn't much data, moving it into
> the file system might be an even better solution.
>
> Russ

Could anyone show me how to do such a thing? (tables + relations --> filesys)

Thanks,
--pac

Re: [9fans] Getting started in Plan9 - help

[Boyd Roberts]

dmr@plan9.bell-labs.com wrote:
> True enough; that said, I retain considerable fondness for
> the history mechanism of later research Unix systems. It
> was adapted from earlier U of Toronto versions, I believe,
> revived by Rob, and later put away again.

Yes, I liked it too, so much so I hacked the SysVr2 shell
to do it and got it to do 'whatis', chucked out 'type'
and added 'builtin'.

Then all I needed was =.c

    http://home.fr.inter.net/boyd/code/=.c

[hmm... that code is old; K&R C]

I believe the code was re-hacked for the unix rc and
fixed a bug with smashing tabs into spaces when you
typed # [delete char] inside a tab.  I could never
bring myself to break the simplicity of the editing
loop to fix that bug.

I also had access to the 8th and 9th Edition sources
to =.c and saw what contorsions it went to to fix the
bug I didn't want to fix.

I assume rob (or someone) ripped the code out of qed to
write =.c

Re: [9fans] Getting started in Plan9 - help

[ozan s yigit]

schwartz@bio.cse.psu.edu (Scott Schwartz) writes:

> used to use a unix thing called "rk"; a markov chain style thing that
> continuously prompted you with a line or two of predicted input.

it is "reactive keyboard" and i believe was a thesis work at university
of calgary, by Darragh under Witten. i'm sure a web search would still turn
up pointers. there is a book about it, not sure if still in print. the
interface was interesting in trying to accomodate disabled people to
interact with command interfaces by predictive completion.

oz
--
www.cs.yorku.ca/~oz	 | if you couldn't find any weirdness, maybe
york u. computer science | we'll just have to make some!   -- hobbes

Re: [9fans] Getting started in Plan9 - help

[Douglas A. Gwyn]

anothy@cosym.net wrote:
> also, as noted earlier, it's not at all clear what the
> "conventional" functionality of del is. on most of my
> Unix systems, it produces ^? - far less useful than the
> interupt function.

On Unix systems since around 7th Edition, one can bind any input
character to the INTR function, and the DEL character was the
default (somewhat confused by Berkeleyites who tried to change
to DEC OS conventions: ^C -> INTR, DEL -> char-erase, etc.).
^? is just a way of representing the ASCII DEL character using
printable notation; the actual character is still DEL (0x7F).

Re: [9fans] Getting started in Plan9 - help

[okamoto]

>I follow this list and browse the documentation in order
>to gain enlightenment and in the hope of gleaning useful ideas for my daily
>work.

By such an attitude, I believe you won't see the problem (if any) the designers
may be facing now, which must be the hottest topic, but then, you'll see only
the past of them.  :-)

I believe any operating system will be tested and trained most effectively in the
real work area, that is any application software.   Therefore, if we try to use Plan 9
for our own  work, we and the community will face an interesting new problem
to be solved, which will promote Plan 9 further.
Don't you think so?
Or you want Plan 9 to be there as it is without more developement?

Kenji

Re: [9fans] Getting started in Plan9 - help

[dmr]

Murdie observed,

 > If you put the command history editor in the shell, then you can only
 > use it in the shell; if you use another shell from time to time, then
 > you have to learn to use that shell's (different) history mechanism.
 > It's far better to use a single, general, command history mechanism
 > provided by your terminal emulator or Acme (which is so more than a
 > terminal emulator). There is a slight loss from the shell and the
 > command history editor being separated, I know.

True enough; that said, I retain considerable fondness for
the history mechanism of later research Unix systems. It
was adapted from earlier U of Toronto versions, I believe,
revived by Rob, and later put away again.

The shell appended each executed command (not in a script)
to a $HISTORY file.

The '=' command (not builtin) searched backwards in the
file for a last matching pattern, and reissued it.
E.g.

	=

redoes the previous command.

	= ls

finds the last shell command beginning with ls.
Text substitution could be accomplished with further
arguments.

The == variant allowed interactive editing.

	Dennis

Re: [9fans] Getting started in Plan9 - help

[Quinn Dunkan]

> I think it might be neat to build a small number
> of little programs that parse and reemit sql queries.
> Thus you could use normal shell pipelines to do database
> queries but since you're passing _queries_ and not _output_
> through the pipes, you still reap the benefits of
> query optimization.  Just like the old /acme/edit scripts.

It's not SQL, but:

http://www.linux.it/~carlos/nosql/

Re: [9fans] Getting started in Plan9 - help

[Andrew Simmons]

>- Why arent there more applications and more developers interested
>in developing for plan9
>

In my case, apart from having to meet deadlines in the Windows world, and
being too cheap to buy a three-button mouse, what rather put me off
seriously pursuing Plan 9 for actual use is a comment some one, I think Mr
Presotto, made a while back to the effect that Plan 9 is a research system,
and that its developers are not very interested in backward compatibility
and only mildly in a large user community. While a perfectly valid goal, I
would have thought it unlikely to attract many people into developing for
Plan 9. As it is, I follow this list and browse the documentation in order
to gain enlightenment and in the hope of gleaning useful ideas for my daily
work.

Re: [9fans] Getting started in Plan9 - help

[erik quanstrom]

along this line, it would be especially cool if the database
engine could export into the filesystem pointers to working
sets, closures, etc. this would allow one to write the
functional equivalent of a stored procedure without any
specific language support in the dtaabase. the place i
always thought this idea fell apart was the other reason
that stored procedures were invented -- so stuff could run
inside the server without the data or the command making a
round trip between client and server. believe me, when you're
walking a million records, it makes a big difference.

the problem is that unless you're really after performance
with insanely-large datasets, a sql-based database just isn't
worth the pain of dealing with that which does not play nicely
with others.

Re: [9fans] Getting started in Plan9 - help

[anothy]

// ...i cant seem to be able to change it...

in my experience (always?), aux/vga is unable to change the
resolution of a display once it's first been brought up. to get
your display to boot in a different resolution, change the
definition of vgasize in plan9.ini (which generally lives in a
FAT partition or on a floppy; more on that in a minute).

Acessibility:

Plan 9 does not include a web browser or other html
interpreter; look to Vita Nuova's freely-available Inferno
distribution, which includes Charon.

man dossrv(4) for a description of the a:, c:, and 9fat:
scripts which will facilitate, respectivly, access to a floppy
drive, your primary FAT partition, and your plan9 FAT
partition, where configuration information like plan9.ini
is commonly stored.

for example, to edit my plan9.ini, i do this:
	9fat:
	acme -c1 /n/9fat/plan9.ini
	unmount /n/9fat

Shell:

to find a file named "myfile" in /usr/anothy, i do:
	`{du -a /usr/anothy | awk '{print $2}'} | grep myfile
this sort of construction works well for all manner of
tools (although i have on occasion missed find's easy
methods of finding files based on greater/less-than
comparisons). also try
	grep some_string `{du -a /usr/anothy | awk '{print $2}'}
for finding "some_string" in a set of files.

rc has no command completion or history. the design
rational is essentially that this functionality properly
belongs in the windowing system, not the shell, so that
the solution becomes general: it can be applied to every
program that looks for input. while not having history
(and to a lesser extent command completion) took me a
bit of getting used to comming from unix, i do not miss
it, and now strongly prefer the Plan 9 way of giving me
such things.

Keys:

plan 9 terminals are much more mouse-driven than
unix terminals. the current UI basically mandates that
you become comfortable using the mouse alot. the end
result is a much more efficient system - the mouse is
much better at many tasks than moving things around
with a keyboard, even when it doesn't feel like it (there
was a good report on this linked here earlier, but i failed
to make a note of it; anyone?). this is not always true
(folks with various disabilities or motor imparment
being the primary counter-example), but is true the
overwhelming majority of the time, and the system is
designed with that in mind.

the interupt functionality could be moved off del, and
onto something else, like the pause/break key. but
keyboards vary widely, and a replacement would have
to be well-chosen. on most smaller keyboards (like on
laptops), del is a key, while pause/break is often a
pseudo-key.

also, as noted earlier, it's not at all clear what the
"conventional" functionality of del is. on most of my
Unix systems, it produces ^? - far less useful than the
interupt function.

General:

i'm not sure what you're asking WRT plan9 default
boot being "safe" for other OSes. the bootloader plan
9 will install is not a multi-booter - it only finds the
first partition marked "active" (or can there only ever
be one?) and runs that. it will not, however, touch
any data living on other partitions.

there has been wide speculation on why Plan 9 has
not gained greater support than it has. ideas on
this topic vary widely. my own theory is something
along the lines of plan 9 is just too much: it requires
people to understand (or be willing to learn) many
new concepts all at once to really "get it". others
believe it has something to do with licensing terms,
or the attitude of people on 9fans and comp.os.plan9.
there's also the old chicken-and-egg problem: users
aren't interested because of the comparitave lack of
apps, and app developers arn't interested because of
the lack of users. but that's all just speculation.
ア

Re: [9fans] Getting started in Plan9 - help

[Matt H]

On Mon, 21 Jan 2002 15:39:45 -0500
"Russ Cox" <rsc@plan9.bell-labs.com> wrote:
> Sql/p

I've begun fiddling with something like that. I began a postgres client, I
was going to return the table as a fixed width sparse matrix in one file
with a file server and also present the same data as a directory tree.

I'm using the current website project as a way to find out the issues of
filesystem storage to make sure I'm not totally barking up the wrong tree

Can't seem to get the time to work on my actual idea.

M

Re: [9fans] Getting started in Plan9 - help

[Russ Cox]

I think it might be neat to build a small number
of little programs that parse and reemit sql queries.
Thus you could use normal shell pipelines to do database
queries but since you're passing _queries_ and not _output_
through the pipes, you still reap the benefits of
query optimization.  Just like the old /acme/edit scripts.

I don't remember enough of the SQL syntax to
reparse your query into an exact shell pipeline, but
the idea is you could do something like

	sql/grep 'Car%' pages.title | sql/join section pages.idx |
	sql/p

Sql/p would behave like the acme p, actually executing
the query to produce output.

Of course, if there isn't much data, moving it into
the file system might be an even better solution.

Russ

Re: [9fans] Getting started in Plan9 - help

[presotto]

To me, dynamic libraries are like clowns.  In both
cases, I was fightened by one at an impressionable
age and have loathed them ever since (Bozo tripped
on his microphone and fell on be when I was 3, Multics
got me at college).

It's a hard fear to overcome.  Putting something into
a dynamic library instead of into a server or the kernel
is oven more efficient.  However, by its very nature,
the interface is less constrained.  As a result, the
visibility of changes and the amount of ratcheting due
to new applications dragging in new, and often incompatible,
versions of the libraries is higher than changes in
the other 2 places.  It's too tempting to add a few
more parameters here and there and to make internal
details externally visible.

I can't really say that they're evil, only that I'ld
prefer not to use them.

Re: [9fans] Getting started in Plan9 - help

[Matt H]

On Mon, 21 Jan 2002 14:42:04 -0500
"David Gordon Hogan" <dhog@plan9.bell-labs.com> wrote:

> Standard practice in the GNU world.  But is it _good_ practice?

_hopefully_ Russ was being sarcastic

> I had a lot of ``fun'' getting gdb to run on Plan 9 because of this...

There is a lot to be said for a common way of getting text from the kb.

Especially when it removes the burden from application programmers from having to deal with user input in their interactive cli applications.

Of course a better way is to remove the interactive part and just expose the interface

So many times I've typed ls and hit return in MySQL console to get a list of tables instead of "show tables"
Every time I say to myself "if this was plan9 ls would work, and work better" because I could use the usual tools on the output

in fact yesterday I did a LOT of mysql processing using the shell and piping the output into awk & sed to make new queries, something you just can't do in the mysql console

The downside is that I have to pass the username & password on the command line too so it starts to get a bit unwieldy :

mysql -r -B -e "select pages.idx, pages.title, pages2.idx as idx2, pages2.title as title2 from pages left join pages as pages2 on pages.section=pages2.idx where pages2.title like'Car%';" -u USER --password='XXXPASSWORDXXXX' content | sed 's/+//' | awk 'BEGIN {FS="\t"} {printf "mysql -r -B -e \"select body from pageblock where pagesidx=%s ; \" -u USER --password='XXXPASSWORDXXXX' content | tr -d \"\\r\" | sed \"s/^body//\" > \"%s.txt\"\n",
$1, $2}'

I've taken Boyd's advice from ages ago that SQL is mostly pointless and I'm moving my tables into the file system and guess what, the code in my CGI is actually shrinking!

I'm losing some type checking and conforming to legal filenames gets in the way sometimes but on the whole I'm satisfied.

I'm hoping to eventually move the whole app across to plan9 and I'm interested in the fact that by using /n/dump I can backtrack to old data, see where the most activity is and stuff like that.

M

Re: [9fans] Getting started in Plan9 - help

[James A. Robinson]

> Only problem with putting it in rio is that you don't have
> quite enough information there.  To do intelligent history
> and completion, you really want to know a few things like
> what program you're doing it for, what their current directory
> is, what their environment variables are, and so on.

Why wouldn't it be a history server which sits inbetween
the keyboard and the rio process?  Records everything sent
in, intercepts whatever you use for commands in order to
print bits of a log back to rio?

Jim

Re: [9fans] Getting started in Plan9 - help

[Mike Haertel]

>> If you put the command history editor in the shell, then you can only
>> use it in the shell; if you use another shell from time to time, then
>
>Not if you put it in a library and link the library
>against every single program that could ever possibly
>want to read input!

Or better yet, put it in rio.  Then every program gets the
feature, and nobody has to remember to use the funny library.

Plus, libraries have some troubles of their own..

* If static libraries, then you have to relink everything
  every time you change the library.

* If dynamic linked libraries, you have taken the first step
  on the slippery slope to the land of gratuitious complexity,
  where even a simple "hello, world" C program can execute
  100000 instructions before it actually prints anything.
  (No, I'm not kidding.)  For some reason, people seem to think
  adding features in dynamic libraries is "free".

Only problem with putting it in rio is that you don't have
quite enough information there.  To do intelligent history
and completion, you really want to know a few things like
what program you're doing it for, what their current directory
is, what their environment variables are, and so on.

Re: [9fans] Getting started in Plan9 - help

[David Gordon Hogan]

> > If you put the command history editor in the shell, then you can only
> > use it in the shell; if you use another shell from time to time, then
>
> Not if you put it in a library and link the library
> against every single program that could ever possibly
> want to read input!

Standard practice in the GNU world.  But is it _good_ practice?

I had a lot of ``fun'' getting gdb to run on Plan 9 because of this...

Re: [9fans] Getting started in Plan9 - help

[Russ Cox]

> If you put the command history editor in the shell, then you can only
> use it in the shell; if you use another shell from time to time, then

Not if you put it in a library and link the library
against every single program that could ever possibly
want to read input!

Re: [9fans] Getting started in Plan9 - help

[John Murdie]

> - Why doesnt/Can rc have autocomplete and filename completion as in
> bash ? This has become so neccessary.

If you put the command history editor in the shell, then you can only
use it in the shell; if you use another shell from time to time, then
you have to learn to use that shell's (different) history mechanism.
It's far better to use a single, general, command history mechanism
provided by your terminal emulator or Acme (which is so more than a
terminal emulator). There is a slight loss from the shell and the
command history editor being separated, I know.

Incidentally, I hate command completion predictors; they remember my
typing mistakes days, weeks or months later, either hesitating to show
me the full, correct, command because of my previous mistake or, worse,
confidently complete my command with the mistake!
--

John A. Murdie
Experimental Officer (Software)
Department of Computer Science
University of York
England

Re: [9fans] Getting started in Plan9 - help

[Boyd Roberts]

steve.simon@snellwilcox.com wrote:
> I think that input prediction, if done well, is a beautiful feature, and
> one that would fit very well with acme, or maybe as a kind of plumbing. I
> used to use a unix thing called "rk"; a markov chain style thing that
> continuously prompted you with a line or two of predicted input. You
> used the arrow keys or tab or ctrl-m to accept the next char/word/line
> of the prediction. It was uncannily good.

In Japan they have a similar style of thing for typing Kanji.  The
pronounciation stem is typed (in kana) and you flip through a list of
choices.  If it's smart it should remember the kanji-frequencies on
a per-user basis.  I never knew if the ones I saw ever did.

Re: [9fans] Getting started in Plan9 - help

[Boyd Roberts]

Matt H wrote:
> > - Why doesnt/Can rc have autocomplete and filename completion as in
> > bash ? This has become so neccessary.
> yes, well, you see plan9 is more mouse driven. eventually you'll probably end up with Acme as much your "shell" as anything, and you'll find auto complete is unneccessary.
> But you're right, it is a nice feature of the bash shell but then there are soooo many goddam directories on a Linux/FreeBSD box and auto complete is Bash's way of trying to alleviate the pain. If you miss it too much I'm sure you could just write a shell script to monitor /dev/cons for tabs, and echo the stuff into /dev/cons.
> Personally, I do prefer having the screen as free form is plan9's is. The shell is more than the commands you can type, it's where you can type them.

I remember the major flamewar over whether Byron's unix implementation of rc
should do this;  I was in the 'no way' camp.  The result was that you could
conditionally compile in that readline trash.  You could probably pick it out
and stick into Plan 9's rc if you wanted to, but Plan 9 is not unix.  It has
much better ways to do things.

I guess another way to do it is to use pipefile.  One of the Kenji's (iirc)
did this for japanese input -- now there's a problem for you.

As for Latin-1: "Fco. J. Ballesteros" <nemo@plan9.escet.urjc.es> has volunteered
to clean up what I did late last year (I'm too busy).  If anyone wants it I'll
send it on or put it on a web page somewhere.  I think the only problem is the
caps-lock/ctrl key swap.

Re: [9fans] Getting started in Plan9 - help

[cej]

Rosh,

you can find some stupid scripts, including "find", at
http://cejchan.gli.cas.cz/plan9

Cheers,
--pac

Re: [9fans] Getting started in Plan9 - help

[Scott Schwartz]

| yes, well, you see plan9 is more mouse driven. eventually you'll
| probably end up with Acme as much your "shell" as anything, and you'll
| find auto complete is unneccessary.

I think that input prediction, if done well, is a beautiful feature, and
one that would fit very well with acme, or maybe as a kind of plumbing. I
used to use a unix thing called "rk"; a markov chain style thing that
continuously prompted you with a line or two of predicted input. You
used the arrow keys or tab or ctrl-m to accept the next char/word/line
of the prediction. It was uncannily good. A lot of command line stuff is
very repetative, and anyone who's seen Rob's fake usenet postings can
see how good this kind of thing is for email. One of these days I'll
get around to hacking it into acme, maybe.

| > - Unless is it part of a grander plan (no pun intended), can we move
| > the process interrupt key from Del to something else and have the
| > conventional functionality of del back ?

Especially since PC keyboards have an actual "break" key to use.

Re: [9fans] Getting started in Plan9 - help

[William Josephson]

On Mon, Jan 21, 2002 at 01:32:35AM +0530, Roshan James wrote:

> - Why doesnt/Can rc have autocomplete and filename completion as in
> bash ? This has become so neccessary.

binding everything on to /bin mostly remove the need for this.
If you haven't done so already, I would suggest grabbing the
various shell scripts and C programs from Russ Cox's web
page at www.eecs.harvard.edu/~rsc.  " and "" are very useful
in conjunction with the mouse.

 -WJ

Re: [9fans] Getting started in Plan9 - help

[rob pike]

> - Is there a place where the uses of directories the std file system
> heirarchy is discussed, esp /n ?

namespace(4).  i admit that the distinction between /n and /mnt is
blurred. the original intent was /n was remote machines, while /mnt
was local services.  a: etc. should probably have been in /mnt.
ditto the blurring for /lib and /sys/lib.  /lib was supposed to be
a library: dictionaries, etc, while /sys/lib was supposed to be supporting
stuff like troff macros.   but we have what we have. see namespace(4).

-rob

Re: [9fans] Getting started in Plan9 - help

[William S.]

I can answer this one:

step one: (at the prompt type) a:
step two: cd /n/a:

Bill
Amsterdam, NL

On Mon, Jan 21, 2002 at 01:32:35AM +0530, Roshan James wrote:
<<snip>>
>
>    - How can i access the floppy a: ? /n/a: exists but shows no files.
<<snip>>

Re: [9fans] Getting started in Plan9 - help

[Matt H]

Hi,

here's my set of slightly flippant answers

> - How can I read a couple of html docs in Plan9 ?
install inferno and use the netscape 3 hybrid Charon
I bet you can't wait :)

Web browsing it's plan9's end user pitfall.
No browser, not even text only (unless you count downloading & stripping the html tags text only)

> - How can I find/search for a file in Plan9 ? the usual find /|grep xxx
> does not exist here, what is the equivalent ?
du I think is your best bet

it's better still to learn where everything is :)

luckily there aren't 5 different directories where programs hide (well there can be but...)
all the executables show themselves in /bin which is a union of the directories where executables live if you see what I mean. There's aren't that many, have a look through them all, you'll remember easily enough.

> - Why doesnt/Can rc have autocomplete and filename completion as in
> bash ? This has become so neccessary.
yes, well, you see plan9 is more mouse driven. eventually you'll probably end up with Acme as much your "shell" as anything, and you'll find auto complete is unneccessary.
But you're right, it is a nice feature of the bash shell but then there are soooo many goddam directories on a Linux/FreeBSD box and auto complete is Bash's way of trying to alleviate the pain. If you miss it too much I'm sure you could just write a shell script to monitor /dev/cons for tabs, and echo the stuff into /dev/cons.
Personally, I do prefer having the screen as free form is plan9's is. The shell is more than the commands you can type, it's where you can type them.

> - Why cant the left/right arrow keys+home+end keys move the cursor,
> it is really difficult to edit something by placing the cursor there
> with the mouse.
That's what I said and I still get the urge to say it out loud. They told me I'd get used to it and you know what, I haven't. I'd even settle for Ctrl-J. But when I'm sat at a different terminal I still end up saying "I wish I was using Acme".

> - Unless is it part of a grander plan (no pun intended), can we move
> the process interrupt key from Del to something else and have the
> conventional functionality of del back ?
It depends who's conventions.

> - Why arent there more applications and more developers interested
> in developing for plan9 ?
file name completion

> This is a lesson that could learned from the Win32's,
> if you want the OS to grow, you have to get people comfortable with
> it very fast. I think we can make that.
After ten years of Windows I'm not sure people are comfortable with it.
It's clunky, crashes without explanation, brittle to end user fiddling, repeatedly exposes remote root exploits, is expensive, closed source. I need not go on.

> One Ring to rule them all, One Ring to find them,
> One Ring to bring them all and in the darkness bind them.
> (Lord of the Rings)
Arntcha sick of those mobiles phones yet?

Matt

[9fans] Getting started in Plan9 - help

[Roshan James]

[-- Attachment #1: Type: text/plain, Size: 3665 bytes --]

Getting started in Plan9
-------------------------    
Its been a little over a week since i got my Plan9 working and 
I still seem to be in tourist mode.Lots of questions and 
a few suggestions:
(I promise I have tried to answer these for myself before
before I am ask them)

It would be great if we have a school boy style step-by-step 
getting-off-the-ground tour of plan9, maybe somewhere in the
wiki. I would be glad to do this, if i knew enough.

Graphics
-----------
- I am working with an S3 Trio 64v2 card, the install floppy
gave me 800*600 res,but after installation i am on 640*480 and
i cant seem to be able to change it
aux/vga -l 800x600x8
gives me
'Warning (BUG) : redefinition of aperture does not change 
s3screen segment.'
in a black background in the sentre of the screen and an error
message that reads 
'aux/vga: vgactlw: <size 800x600x8 m8>: vga already configured'
in the console window. it is a low end card but I believe that 
I did have a higher res through the boot disk so it should be 
possible here too. how can i change to a higher res ?

- If plan9 is booted through xosl in 640*480 res,plan9 graphics
display ends up corrupt. the bootloader does switch to text mode
before the OS is booted. anyother resolution or a text mode boot
loader does not seem to have a problem. 
The right quarter of the screen (approx) seems to be a duplicate
of the band of the screen display between in the left part. (bad
description i know). Anyway to fix this ?


Acessibility
-------------
- How can I read a couple of html docs in Plan9 ?
- Is there a place where the uses of directories the std file system
heirarchy is discussed, esp /n ? 
- /n/c: exists, how can i access the extended partitions ?
- How can i access the floppy a: ? /n/a: exists but shows no files.
- How can i access the extended windows partitions ? 
- Problem with accessing C: File operations to /n/c: causes a problem
'%mkdir /n/c:/testdir'
'mkdir: cant create /n/c:/testdir: write to hungup channel'
also a black background error message comes (is there a generic name
for these messages ?)
'dossrv 45: suicide: sys: trap fault read addr=0xb pc=0x00004757'
help ?

Shell
------ 
- How can I find/search for a file in Plan9 ? the usual find /|grep xxx 
does not exist here, what is the equivalent ?
- Why doesnt/Can rc have autocomplete and filename completion as in 
bash ? This has become so neccessary.

Keys
-----
- Why cant the left/right arrow keys+home+end keys move the cursor,
it is really difficult to edit something by placing the cursor there
with the mouse. 
- Unless is it part of a grander plan (no pun intended), can we move 
the process interrupt key from Del to something else and have the 
conventional functionality of del back ?

General
-------
- Is the option of plan9 default boot in bootsetup (during install)
safe for other OSes that exist on the system ?
- Why arent there more applications and more developers interested
in developing for plan9 ? 

Russ, I think it would kill you to keep answering all the newbie 
questions. Russ, Imel, Thanks for all the help you have been. I 
think the Plan9 faq needs updation with some of the more generic 
questions here. This is a lesson that could learned from the Win32's, 
if you want the OS to grow, you have to get people comfortable with 
it very fast. I think we can make that happen.

Rosh.

-------------------------------------------------------------------------------------
One Ring to rule them all, One Ring to find them,
One Ring to bring them all and in the darkness bind them.
(Lord of the Rings)

[-- Attachment #2: Type: text/html, Size: 4834 bytes --]