9fans - fans of the OS Plan 9 from Bell Labs
 help / color / mirror / Atom feed
* [9fans] control-F completion question
@ 2008-10-03 16:39 Rudolf Sykora
  2008-10-03 16:52 ` andrey mirtchovski
                   ` (3 more replies)
  0 siblings, 4 replies; 8+ messages in thread
From: Rudolf Sykora @ 2008-10-03 16:39 UTC (permalink / raw)
  To: Fans of the OS Plan 9 from Bell Labs

[-- Attachment #1: Type: text/plain, Size: 1037 bytes --]

Hello everybody!

If I understand it right ^f (or an 'ins' key) are taken care of by rio and
thus the success of completion is essentially dependent on the namespace rio
is using. This namespace is created when rio is started, usually right after
a computer start. When the namespaces of individual windows are changed, e.
g. by binding some remote filesystems, ^f can't handle those new files
(since the rio namespace stays intact). Even though I may understand the
reason (i.e. what I have just said) I find it rather irritating, having
maybe the whole space I work with out of reach of ^f. Is there any help with
that? Couldn't it be somehow achieved that ^f worked 'better'? (Not saying
rc should take care of it, it probably should not; but what about if it were
somehow connected with the individual windows? -- I don't know, it may not
be possible, just asking. Having to always write 'lc' is somewhat ...).
Starting a bunch of several rios can help it. But is that a right way to
go?

Thanks for answers.
Ruda

[-- Attachment #2: Type: text/html, Size: 1104 bytes --]

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [9fans] control-F completion question
  2008-10-03 16:39 [9fans] control-F completion question Rudolf Sykora
@ 2008-10-03 16:52 ` andrey mirtchovski
  2008-10-03 17:11   ` Rudolf Sykora
  2008-10-03 16:53 ` Nathaniel W Filardo
                   ` (2 subsequent siblings)
  3 siblings, 1 reply; 8+ messages in thread
From: andrey mirtchovski @ 2008-10-03 16:52 UTC (permalink / raw)
  To: Fans of the OS Plan 9 from Bell Labs

i sometimes wonder whether the ^F completion stuff wasn't left this
way on purpose, as if to illustrate the futility of (essentially) a
single-node solution in the presence of distributed environments.

to solve what you perceive to be a problem you must ask yourself: what
is the one thing that is fully aware of the current namespace? it's
obviously not rio: it simply juggles windows with shells in them. is
it the shell? putting the completion in the shell itself would work
for plain terminals, but wouldn't work for rio. is it the kernel?
would you bother adding to the kernel something as silly as command
completion? how about completing across a network?

all difficult questions for a silly problem ;)

On Fri, Oct 3, 2008 at 10:39 AM, Rudolf Sykora <rudolf.sykora@gmail.com> wrote:
> Hello everybody!
>
> If I understand it right ^f (or an 'ins' key) are taken care of by rio and
> thus the success of completion is essentially dependent on the namespace rio
> is using. This namespace is created when rio is started, usually right after
> a computer start. When the namespaces of individual windows are changed, e.
> g. by binding some remote filesystems, ^f can't handle those new files
> (since the rio namespace stays intact). Even though I may understand the
> reason (i.e. what I have just said) I find it rather irritating, having
> maybe the whole space I work with out of reach of ^f. Is there any help with
> that? Couldn't it be somehow achieved that ^f worked 'better'? (Not saying
> rc should take care of it, it probably should not; but what about if it were
> somehow connected with the individual windows? -- I don't know, it may not
> be possible, just asking. Having to always write 'lc' is somewhat ...).
> Starting a bunch of several rios can help it. But is that a right way to
> go?
>
> Thanks for answers.
> Ruda
>



^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [9fans] control-F completion question
  2008-10-03 16:39 [9fans] control-F completion question Rudolf Sykora
  2008-10-03 16:52 ` andrey mirtchovski
@ 2008-10-03 16:53 ` Nathaniel W Filardo
  2008-10-03 17:32 ` erik quanstrom
  2008-10-03 18:42 ` Steve Simon
  3 siblings, 0 replies; 8+ messages in thread
From: Nathaniel W Filardo @ 2008-10-03 16:53 UTC (permalink / raw)
  To: Fans of the OS Plan 9 from Bell Labs

[-- Attachment #1: Type: text/plain, Size: 1614 bytes --]

On Fri, Oct 03, 2008 at 06:39:17PM +0200, Rudolf Sykora wrote:
> Hello everybody!
> 
> If I understand it right ^f (or an 'ins' key) are taken care of by rio and
> thus the success of completion is essentially dependent on the namespace rio
> is using. This namespace is created when rio is started, usually right after
> a computer start. When the namespaces of individual windows are changed, e.
> g. by binding some remote filesystems, ^f can't handle those new files
> (since the rio namespace stays intact). Even though I may understand the
> reason (i.e. what I have just said) I find it rather irritating, having
> maybe the whole space I work with out of reach of ^f. Is there any help with
> that? Couldn't it be somehow achieved that ^f worked 'better'? (Not saying
> rc should take care of it, it probably should not; but what about if it were
> somehow connected with the individual windows? -- I don't know, it may not
> be possible, just asking. Having to always write 'lc' is somewhat ...).
> Starting a bunch of several rios can help it. But is that a right way to
> go?
> 
> Thanks for answers.
> Ruda

This was the issue that prompted my experimental work for cross-namespace
visibility (see
https://wiki.ietfng.org/pub/Plan9/KernelInternals/CrossNamespaceWalkProject
if you're curious).  This experiment is probably not the Right Way to do
this, but it may be food for thought.

Another alternative would be to spawn an exportfs "next to" the rc inside
the window, and have rio use that to find completions.  Not terribly
pleasant, but might be workable.

--nwf;

[-- Attachment #2: Type: application/pgp-signature, Size: 204 bytes --]

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [9fans] control-F completion question
  2008-10-03 16:52 ` andrey mirtchovski
@ 2008-10-03 17:11   ` Rudolf Sykora
  2008-10-03 17:34     ` erik quanstrom
  2008-10-03 17:34     ` Federico G. Benavento
  0 siblings, 2 replies; 8+ messages in thread
From: Rudolf Sykora @ 2008-10-03 17:11 UTC (permalink / raw)
  To: Fans of the OS Plan 9 from Bell Labs

[-- Attachment #1: Type: text/plain, Size: 1314 bytes --]

2008/10/3 andrey mirtchovski <mirtchovski@gmail.com>

> i sometimes wonder whether the ^F completion stuff wasn't left this
> way on purpose, as if to illustrate the futility of (essentially) a
> single-node solution in the presence of distributed environments.
>
> to solve what you perceive to be a problem you must ask yourself: what
> is the one thing that is fully aware of the current namespace? it's
> obviously not rio: it simply juggles windows with shells in them. is
> it the shell? putting the completion in the shell itself would work
> for plain terminals, but wouldn't work for rio. is it the kernel?
> would you bother adding to the kernel something as silly as command
> completion? how about completing across a network?
>
> all difficult questions for a silly problem ;)
>

I might again be wrong, but I thought the reason why the completion is not
in the rc is partially thanks to the fact, that rc doesn't know anything you
type until you press 'enter'. Only rio knows. (But all this is based only on
my possibly broken opinion...)
But yes, the best would be (for me and I guess for everybody, since we
usually speak to rc when we need the completion) to be as close to the rc's
namespace as possible.
Correct me, please, if I am completely off the road... :)

Ruda

[-- Attachment #2: Type: text/html, Size: 1725 bytes --]

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [9fans] control-F completion question
  2008-10-03 16:39 [9fans] control-F completion question Rudolf Sykora
  2008-10-03 16:52 ` andrey mirtchovski
  2008-10-03 16:53 ` Nathaniel W Filardo
@ 2008-10-03 17:32 ` erik quanstrom
  2008-10-03 18:42 ` Steve Simon
  3 siblings, 0 replies; 8+ messages in thread
From: erik quanstrom @ 2008-10-03 17:32 UTC (permalink / raw)
  To: 9fans

> Hello everybody!
>
> If I understand it right ^f (or an 'ins' key) are taken care of by rio and
> thus the success of completion is essentially dependent on the namespace rio
> is using. This namespace is created when rio is started, usually right after
> a computer start. When the namespaces of individual windows are changed, e.
> g. by binding some remote filesystems, ^f can't handle those new files
> (since the rio namespace stays intact). Even though I may understand the
> reason (i.e. what I have just said) I find it rather irritating, having
> maybe the whole space I work with out of reach of ^f. Is there any help with
> that? Couldn't it be somehow achieved that ^f worked 'better'? (Not saying
> rc should take care of it, it probably should not; but what about if it were
> somehow connected with the individual windows? -- I don't know, it may not
> be possible, just asking. Having to always write 'lc' is somewhat ...).
> Starting a bunch of several rios can help it. But is that a right way to
> go?
>
> Thanks for answers.
> Ruda

you can accomplish this simply by using the plumber which,
by convention, shares a namespace with rio.  suppose you
want insert to work with sources.  then plumbing the string
	Local 9fs sources
will allow insert to wor.  alternatively, one might wish
to import sources with an aan connection before starting
rio.  this works pretty well for me since most of the time
it's fairly easy to anticipate the stuff you're going to add
to the namespace.

- erik




^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [9fans] control-F completion question
  2008-10-03 17:11   ` Rudolf Sykora
  2008-10-03 17:34     ` erik quanstrom
@ 2008-10-03 17:34     ` Federico G. Benavento
  1 sibling, 0 replies; 8+ messages in thread
From: Federico G. Benavento @ 2008-10-03 17:34 UTC (permalink / raw)
  To: Fans of the OS Plan 9 from Bell Labs

hola,

I got the tip from rog's post to affect rio's ns,
plumber's actually.

http://9fans.net/archive/2005/02/171

On Fri, Oct 3, 2008 at 3:11 PM, Rudolf Sykora <rudolf.sykora@gmail.com> wrote:
>
>
> 2008/10/3 andrey mirtchovski <mirtchovski@gmail.com>
>>
>> i sometimes wonder whether the ^F completion stuff wasn't left this
>> way on purpose, as if to illustrate the futility of (essentially) a
>> single-node solution in the presence of distributed environments.
>>
>> to solve what you perceive to be a problem you must ask yourself: what
>> is the one thing that is fully aware of the current namespace? it's
>> obviously not rio: it simply juggles windows with shells in them. is
>> it the shell? putting the completion in the shell itself would work
>> for plain terminals, but wouldn't work for rio. is it the kernel?
>> would you bother adding to the kernel something as silly as command
>> completion? how about completing across a network?
>>
>> all difficult questions for a silly problem ;)
>
> I might again be wrong, but I thought the reason why the completion is not
> in the rc is partially thanks to the fact, that rc doesn't know anything you
> type until you press 'enter'. Only rio knows. (But all this is based only on
> my possibly broken opinion...)
> But yes, the best would be (for me and I guess for everybody, since we
> usually speak to rc when we need the completion) to be as close to the rc's
> namespace as possible.
> Correct me, please, if I am completely off the road... :)
>
> Ruda
>
>



--
Federico G. Benavento



^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [9fans] control-F completion question
  2008-10-03 17:11   ` Rudolf Sykora
@ 2008-10-03 17:34     ` erik quanstrom
  2008-10-03 17:34     ` Federico G. Benavento
  1 sibling, 0 replies; 8+ messages in thread
From: erik quanstrom @ 2008-10-03 17:34 UTC (permalink / raw)
  To: 9fans

> I might again be wrong, but I thought the reason why the completion is not
> in the rc is partially thanks to the fact, that rc doesn't know anything you
> type until you press 'enter'. Only rio knows. (But all this is based only on
> my possibly broken opinion...)
> But yes, the best would be (for me and I guess for everybody, since we
> usually speak to rc when we need the completion) to be as close to the rc's
> namespace as possible.
> Correct me, please, if I am completely off the road... :)
>
> Ruda

really? i use completion in the acme tag line more often.

- erik




^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [9fans] control-F completion question
  2008-10-03 16:39 [9fans] control-F completion question Rudolf Sykora
                   ` (2 preceding siblings ...)
  2008-10-03 17:32 ` erik quanstrom
@ 2008-10-03 18:42 ` Steve Simon
  3 siblings, 0 replies; 8+ messages in thread
From: Steve Simon @ 2008-10-03 18:42 UTC (permalink / raw)
  To: 9fans

rio does its magic (<ins> expanding filenames) by consulting /dev/wdir,
thus any way you can keep wdir up to date will allow <ins> to work,
(and the plumber to find files).

You can even use u9fs (or sftpfs) to connect to a unix box and ssh to
connect to a remote unix host, and plumb files (or use <ins> to expand
file names), this parses the current directory embedded in the unix
host's prompt  - see rwd(1).

I have a similar system for accessing windows servers and windows workstations,
though I use my own server on the workstation rather than openssh.

-Steve



^ permalink raw reply	[flat|nested] 8+ messages in thread

end of thread, other threads:[~2008-10-03 18:42 UTC | newest]

Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2008-10-03 16:39 [9fans] control-F completion question Rudolf Sykora
2008-10-03 16:52 ` andrey mirtchovski
2008-10-03 17:11   ` Rudolf Sykora
2008-10-03 17:34     ` erik quanstrom
2008-10-03 17:34     ` Federico G. Benavento
2008-10-03 16:53 ` Nathaniel W Filardo
2008-10-03 17:32 ` erik quanstrom
2008-10-03 18:42 ` Steve Simon

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).