how to find out methods for tramp?

how to find out methods for tramp?

[Robert J. Chassell]

Kai just added `tramp' to the latest GNU Emacs CVS.

How do I find out which methods to use for tramp connections?

Using today's CVS snapshot of Mon, 2002 Jun 17  15:19 UTC, 
GNU Emacs 21.3.50.42 (i686-pc-linux-gnu, X toolkit),
started with:

    emacs -q --no-site-file --eval '(blink-cursor-mode 0)'

and 

    (require 'tramp)

I can successfully use tramp to connect remotely to

    fencepost.gnu.org 

with 
    (setq tramp-default-method "smx") 
which is `ssh' with `mimencode'

But I failed to connect to two other machines, to both of which I can
connect using ssh:

My ISP, a Sun box, called `titan'"

    SunOS 5.6 Generic_105181-08 sun4u sparc
    has  uuencode but not mimencode

    I tried smx, sm, su, and tu  and failed with tramp.

With     (setq tramp-default-method "su") 

for example, I was shown a login and the process was immediately
killed:

    <bob@titan> ~ $ 
    Process *tramp/su titan.berkshire.net* killed

    I can connect to titan from an xterm using `ssh'

My sister's box, an i586 

    GNU/Linux, kernel 2.4.18
    has mimencode
    running OpenSSH_3.0.2p1 Debian 1:3.0.2p1-9, 
       SSH protocols 1.5/2.0, OpenSSL 0x0090603f
      and permits remote connections using only protocol 2

    I tried smx and sm and failed with tramp.

    I can connect to this box from an xterm using `ssh -2'

Here are the results for tramp:

With:

    (setq tramp-default-method "smx")
    (setq tramp-su-args nil)

        Protocol major versions differ: 1 vs. 2
        Process *tramp/smx 24.161.120.95* exited abnormally with code 255

With:

    (setq tramp-default-method "smx")
    (setq tramp-su-args -2)

        tramp-open-connection-rsh: Couldn't find remote shell or passwd prompt

With:

    (setq tramp-default-method "sm")
    (setq tramp-su-args -2)

        tramp-open-connection-rsh: Couldn't find remote shell or passwd prompt


Thank you.

-- 
    Robert J. Chassell                  bob@rattlesnake.com
    Rattlesnake Enterprises             http://www.rattlesnake.com

Re: how to find out methods for tramp?

[Andreas Schwab]

"Robert J. Chassell" <bob@rattlesnake.com> writes:

|> My sister's box, an i586 
|> 
|>     GNU/Linux, kernel 2.4.18
|>     has mimencode
|>     running OpenSSH_3.0.2p1 Debian 1:3.0.2p1-9, 
|>        SSH protocols 1.5/2.0, OpenSSL 0x0090603f
|>       and permits remote connections using only protocol 2
|> 
|>     I tried smx and sm and failed with tramp.
|> 
|>     I can connect to this box from an xterm using `ssh -2'

That's probably best solved by an entry in ~/.ssh/config.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Deutschherrnstr. 15-19, D-90429 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: how to find out methods for tramp?

[Miles Bader]

Andreas Schwab <schwab@suse.de> writes:
> |>     I can connect to this box from an xterm using `ssh -2'
> 
> That's probably best solved by an entry in ~/.ssh/config.

[Note, just in case it's not obvious, you can easily have per-host
configuration parameters in ~/.ssh/config; see the docs about the `Host ...'
line in that file.]

-Miles
-- 
Saa, shall we dance?  (from a dance-class advertisement)

Re: how to find out methods for tramp?

[Richard Stallman]

Bob, did you read the Tramp manual to get help for setting this up?
If not, please try that and tell us if it helps you.

To the others: Bob is a very capable user, for a non-hacker.  If he
has trouble doing something, that means it is hard; if Bob doesn't
understand the documentation, it must be quite unclear.  Making
documentation clear to Bob does not guarantee it is good, but that is
a necessary minimum requirement.

Re: how to find out methods for tramp?

[Robert J. Chassell]

   Bob, did you read the Tramp manual to get help for setting this up?
   If not, please try that and tell us if it helps you.

Yes, I did read it.  As I wrote earlier, it described the methods
quite well.  But it did not tell me how to choose the best method
efficiently; only to try each method one at a time until one worked.

There are three issues here:

  * Possible bugs in the code.  Kai asked me, and I have sent him, a
    trace.  (One of the non-working sites is using what my successfull
    SSH-connection-in-an-xterm reports as a possibly `old'
    implementation of SSH; the other requires protocol 2 and Tramp
    fails with any SSH configuration that also permits protocol 1.)

  * Inadequately informative error messages.  When Tramp fails, the
    error messages should tell me more as to why it fails:  for
    example, I can successfully use Tramp to list my home directory on
    fp.gnu.org using the `smx' method using yesterday's CVS snapshot,
    started with: emacs -q --no-site-file --eval '(blink-cursor-mode 0)'
   
    However, when I try the same operation using the `sm' method,
    Tramp fails and tells me only that

        tramp-open-connection-rsh: Wrong type argument: sequencep, 60

    The  *debug tramp/sm fp.gnu.org* provides even less help.

    Neither suggests which of the other 28 or so methods I should use.

  * Confusing documentation.  

    - The existing documentation, which in general I find pretty clear
      is, in this instance somewhat misleading.  It says that `smx'

           ... is useful for users where the normal login shell
           is set up to ask them a number of questions when logging in. 

      In this case, fencepost does not ask me for a number of
      questions.  It asks one question, which is for my password.  So
      if I follow this documentation, I avoid using the `smx' method.
      However, the `smx' method is the method that works.

      Perhaps the problem is that this instance of SSH negotiates a
      series of questions back and forth, and the `sm' method cannot
      handle that.  As a user I have no way of knowing.  After all, as
      I said, to me fencepost asks only one question, which is for my
      password; and the `sm' method is for other SSH connections to
      sites with `mimencode'.

    - Secondly, while the documentation is helpful in saying that `smx'

          ... is also useful for Windows users ...

      since fencepost is, I am pretty sure, a system that can run X,
      which is a windowing system, I am not 100% sure that the
      documentation writer actually means

          `also useful for users running a windowing system such as X or
          Berlin'

      Perhaps the writer means 

          `also useful for users running the windowing system
          distributed by the Microsoft Corporation'.

      The problem here occurs because some people support the
      Microsoft Corporation's efforts to deny that windowing systems
      are a generic application invented by Englebart in the 1960s.
      When they refer to a windowing system, they mean only Microsoft
      Windows.  (The Microsoft Corporation has spent a great deal of
      effort in trying to deny the everyday meaning of the word
      `Windows' in the context of computers, going so far as to obtain
      a trademark on the word.)

      When I see the word `Windows', while I presume that most likely
      the speaker is talking about computer windowing systems in their
      everyday meaning, referring to X and to SmallTalk, and
      Englebart's work, I have to recognize that the term is ambiguous
      and may refer only to the windowing system distributed and
      claimed by the Microsoft Corporation.

      The documentation would be clearer if it distinguished among

            any windowing system, such as X, or that produced by the
            Apple Corporation, or that produced by the Microsoft
            Corporation

            the windowing system produced by the Microsoft Corporation

            the X windowing system

    and so on.  And, since people always shorten terms in frequent
    use, shorting `television' to TV (or, in the UK, to `tely'), it is
    reasonable sometimes to shorten specific references to Macintosh
    Windows, X Windows, or Microsoft Windows.

-- 
    Robert J. Chassell                  bob@rattlesnake.com
    Rattlesnake Enterprises             http://www.rattlesnake.com

Re: how to find out methods for tramp?

[Stefan Monnier]

>           ... is also useful for Windows users ...
> 
>       since fencepost is, I am pretty sure, a system that can run X,
>       which is a windowing system, I am not 100% sure that the
>       documentation writer actually means

I think the use of a capital together with the particular grammatical
construct used makes it clear that this use of the word `windows'
refers to a proper name rather than to the general concept of a window.
I'm surprised that anybody would be confused.

>       Windows.  (The Microsoft Corporation has spent a great deal of
>       effort in trying to deny the everyday meaning of the word
>       `Windows' in the context of computers, going so far as to obtain
>       a trademark on the word.)

I can only agree that this is a serious issue and that maybe we
should try and do something about it (including rewriting the
above sentence, if some other formulation can be found), but that
still doesn't make the above sentence ambiguous in my book.


	Stefan

Re: how to find out methods for tramp?

[Robert J. Chassell]

   I think the use of a capital together with the particular grammatical
   construct used makes it clear that this use of the word `windows'
   refers to a proper name rather than to the general concept of a window.
   I'm surprised that anybody would be confused.

I get confused because many of the people I know refer to `X Windows'
and shorten that phrase to `Windows'.

I also know some people who use Microsoft operating systems.
Sometimes they use the word `Windows' the way the Microsoft marketing
department wants them to, in a manner that denies the existence of any
other windowing system.  This multiple use of the word is what causes
the ambiguity.

-- 
    Robert J. Chassell                  bob@rattlesnake.com
    Rattlesnake Enterprises             http://www.rattlesnake.com

Re: how to find out methods for tramp?

[Stefan Monnier]

>    I think the use of a capital together with the particular grammatical
>    construct used makes it clear that this use of the word `windows'
>    refers to a proper name rather than to the general concept of a window.
>    I'm surprised that anybody would be confused.
> 
> I get confused because many of the people I know refer to `X Windows'
> and shorten that phrase to `Windows'.

Hmm...interesting.  I've always heard it shortened to `X' which
is indeed shorter than `Windows'.  But at least now I understand
your confusion, thanks.


	Stefan

Re: how to find out methods for tramp?

[Robert J. Chassell]

   Hmm...interesting.  I've always heard it shortened to `X' ...

Yes, some people I know say or write that, too.  My experience is that
people say all sorts of things; and sometimes I am confused.

-- 
    Robert J. Chassell                  bob@rattlesnake.com
    Rattlesnake Enterprises             http://www.rattlesnake.com

Re: how to find out methods for tramp?

[Andreas Schwab]

"Robert J. Chassell" <bob@rattlesnake.com> writes:

|>    I think the use of a capital together with the particular grammatical
|>    construct used makes it clear that this use of the word `windows'
|>    refers to a proper name rather than to the general concept of a window.
|>    I'm surprised that anybody would be confused.
|> 
|> I get confused because many of the people I know refer to `X Windows'
|> and shorten that phrase to `Windows'.

Actually, if you ask the X Consortium then `X Windows' is not a proper
name for the `X Window System' (see X(1)).

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Deutschherrnstr. 15-19, D-90429 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: how to find out methods for tramp?

[Robert J. Chassell]

   Actually, if you ask the X Consortium then `X Windows' is not a proper
   name for the `X Window System' (see X(1)).

Yes, very true.  Yet almost everyone I know says `Windows', `X
Windows', or `X' to refer to the X windowing system.

Similarly, people I know have used the term `Xerox' to mean `make a
plain paper copy' on a machine not produced by the Xerox corporation.

-- 
    Robert J. Chassell                  bob@rattlesnake.com
    Rattlesnake Enterprises             http://www.rattlesnake.com

Re: how to find out methods for tramp?

[Andreas Schwab]

"Robert J. Chassell" <bob@rattlesnake.com> writes:

|> Similarly, people I know have used the term `Xerox' to mean `make a
|> plain paper copy' on a machine not produced by the Xerox corporation.

There are quite a few other brand names that are used so commonly that
they became generic names.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Deutschherrnstr. 15-19, D-90429 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: how to find out methods for tramp?

[Richard Stallman]

    Actually, if you ask the X Consortium then `X Windows' is not a proper
    name for the `X Window System' (see X(1)).

I suspect that pressure from Microsoft had to do with making them
state that position.  So I think that by doing this we cater to
Microsoft rather than to the real wishes of the X developers.

(The X Consortium does not exist any more.  The people who maintain X
nowadays, the XFree86 group, told us they don't care about this issue.)

Re: how to find out methods for tramp?

[Per Abrahamsen]

Richard Stallman <rms@gnu.org> writes:

>     Actually, if you ask the X Consortium then `X Windows' is not a proper
>     name for the `X Window System' (see X(1)).
>
> I suspect that pressure from Microsoft had to do with making them
> state that position.  So I think that by doing this we cater to
> Microsoft rather than to the real wishes of the X developers.

That was the official position of the X developers long before
Microsoft had any idea they would eventually produce a window system.

"It is a window system called X, not a system called X windows."

Re: how to find out methods for tramp?

[Miles Bader]

Per Abrahamsen <abraham@dina.kvl.dk> writes:
> That was the official position of the X developers long before
> Microsoft had any idea they would eventually produce a window system.
> 
> "It is a window system called X, not a system called X windows."

... and many (probably a majority of) people have ignored their wish
from the very beginning...

-Miles
-- 
`Life is a boundless sea of bitterness'

Re: how to find out methods for tramp?

[Kai Großjohann]

"Robert J. Chassell" <bob@rattlesnake.com> writes:

> Kai just added `tramp' to the latest GNU Emacs CVS.
>
> How do I find out which methods to use for tramp connections?

The variable tramp-methods contains a complete list of available
methods.

In the Tramp info file, there is a node which describes all those
methods in greater detail, or at least tries to.  Is that node
helpful at all?

> But I failed to connect to two other machines, to both of which I can
> connect using ssh:
>
> My ISP, a Sun box, called `titan'"
>
>     SunOS 5.6 Generic_105181-08 sun4u sparc
>     has  uuencode but not mimencode
>
>     I tried smx, sm, su, and tu  and failed with tramp.

Could you please (setq tramp-debug-buffer t) {this might be the
default value} and start a fresh Emacs, invoke C-x C-f
/[sm/user@host]/path/to/file RET, then send me the *debug tramp/foo*
buffer as well as the *tramp/foo* buffer?

Please type in the remote file name completely, without using
filename completion.  This means that the debug buffer is smaller,
which makes it easier for me to look at it.

You might wish to have a look at the debug buffer to see if it
contains confidential information.  When you delete that, please
indicate where you have deleted stuff.

> My sister's box, an i586 
>
>     GNU/Linux, kernel 2.4.18
>     has mimencode
>     running OpenSSH_3.0.2p1 Debian 1:3.0.2p1-9, 
>        SSH protocols 1.5/2.0, OpenSSL 0x0090603f
>       and permits remote connections using only protocol 2
>
>     I tried smx and sm and failed with tramp.
>
>     I can connect to this box from an xterm using `ssh -2'

I think there are methods sm1 and sm2 which explicitly invoke the
right protocol version.  But I'm not exactly sure what they are doing.

Anyhow, it should be easy to add an entry to tramp-methods which adds
the right args to the "ssh" invocation.

[time passes]

Ah, the sm1 and sm2 methods invoke ssh1 and ssh2, respectively.  Hm.
So maybe we need variants which do "ssh -1" and "ssh -2",
respectively.  What should their names be?

Maybe the methods invoking ssh1 or ssh2 instead of ssh can be deleted
now that ssh1 is getting really old?

kai
-- 
People mountain, people sea, today no see, tomorrow see.  (from Chinese)

Re: how to find out methods for tramp?

[Stefan Monnier]

> Ah, the sm1 and sm2 methods invoke ssh1 and ssh2, respectively.  Hm.
> So maybe we need variants which do "ssh -1" and "ssh -2",
> respectively.  What should their names be?
> 
> Maybe the methods invoking ssh1 or ssh2 instead of ssh can be deleted
> now that ssh1 is getting really old?

They should definitely be removed.  AFAIK ssh1 and ssh2 were only used
with SSH2 (the program that started the SSH2 protocol but which was
not free software or even open source) because it did not support
the SSH1 protocol directly (so as to force people to switch to SSH2
more quickly, although this impression might just be the result of
my paranoia).
OpenSSH never needed any such thing because it started with SSH1 support
and added SSH2 later on in a fully backward compatible way, using the
-1 and -2 arguments.


	Stefan

Re: how to find out methods for tramp?

[Robert J. Chassell]

   > How do I find out which methods to use for tramp connections?

   The variable tramp-methods contains a complete list of available
   methods.

   In the Tramp info file, there is a node ...

Yes, I looked at those.  The are fine explanations.

The problem is, the only way I could choose one method instead of
another was to try a whole bunch of methods.  That is the problem and
is inefficent.

-- 
    Robert J. Chassell                  bob@rattlesnake.com
    Rattlesnake Enterprises             http://www.rattlesnake.com

Re: how to find out methods for tramp?

[Eli Zaretskii]

On Tue, 18 Jun 2002, Robert J. Chassell wrote:

>    In the Tramp info file, there is a node ...
> 
> Yes, I looked at those.  The are fine explanations.
> 
> The problem is, the only way I could choose one method instead of
> another was to try a whole bunch of methods.  That is the problem and
> is inefficent.

This probably means that the Tramp manual should explain the methods more 
than it does now, perhaps from a different angle: it should make it easy 
for a user to choose the suitable method.

One idea is to add a kind of ``cookbook'': a @multitable which lists 
several popular situations and the best Tramp method which matches each 
situation.

Another idea is to discuss the advantages and disadvantages of each 
method, in a way that makes it easier to decide which is the best one for 
you.

Re: how to find out methods for tramp?

[Kai Großjohann]

"Robert J. Chassell" <bob@rattlesnake.com> writes:

> The problem is, the only way I could choose one method instead of
> another was to try a whole bunch of methods.  That is the problem and
> is inefficent.

This is bad.  I agree with Eli and Richard that this means that the
manual needs to be fixed.  Thanks a lot for finding this problem.

Can you say which kind of explanation would have helped you to find
the right method quickly?

Maybe the right way to explain it is to start from the user's
situation.  We should get the user to log in to the remote host from
the command line.  The command used there will tell us something
about which methods are applicable.

Then we could tell the user to issue commands like "type mimencode"
and "type uuencode", as you have used.  This will tell us more about
the applicable methods.

Another thing to say is that people should start with inline methods
and try to get that working first.  If the inline method works but
turns out to be too slow, they can still work on getting an
out-of-band method to work.

kai
-- 
People mountain, people sea, today no see, tomorrow see.  (from Chinese)

Re: how to find out methods for tramp?

[Kim F. Storm]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:

> Maybe the right way to explain it is to start from the user's
> situation.  We should get the user to log in to the remote host from
> the command line.  The command used there will tell us something
> about which methods are applicable.
> 
> Then we could tell the user to issue commands like "type mimencode"
> and "type uuencode", as you have used.  This will tell us more about
> the applicable methods.

Can't you write some code in elisp which does this sort of thing
automatically, and determines the best method to use for a given
system?

> 
> Another thing to say is that people should start with inline methods
> and try to get that working first.  If the inline method works but
> turns out to be too slow, they can still work on getting an
> out-of-band method to work.

The automated detection can try the inline methods first, and only
look for out-of-band methods if none of the inline methods work.

-- 
Kim F. Storm <storm@cua.dk> http://www.cua.dk

Re: how to find out methods for tramp?

[Kai Großjohann]

storm@cua.dk (Kim F. Storm) writes:

> Can't you write some code in elisp which does this sort of thing
> automatically, and determines the best method to use for a given
> system?

Of course it's (in principle) possible to do that.  However, Tramp is
already doing a lot of auto-detection, and I didn't want to add to it
unless absolutely necessary.  (There are a lot of places where things
can go wrong, and so by letting the user choose the connection
method, some of those things are blamed on the user :-)

I want to change Tramp such that it auto-detects more things.  This
will take time, however.

kai
-- 
A large number of young women don't trust men with beards.  (BFBS Radio)

Re: how to find out methods for tramp?

[Robert J. Chassell]

   "Robert J. Chassell" <bob@rattlesnake.com> writes:

   > The problem is, the only way I could choose one method instead of
   > another was to try a whole bunch of methods.  That is the problem and
   > is inefficent.

   This is bad.  I agree with Eli and Richard that this means that the
   manual needs to be fixed.  Thanks a lot for finding this problem.

   Can you say which kind of explanation would have helped you to find
   the right method quickly?

Ideally, Tramp should try additional methods if the method you use
fails (with an option to turn off this feature by setting a variable).

In any case, messages that go to the *Message* buffer should be more
helpful in telling what happened.  They should tell you 

    `mimencode' not found on this machine; try a Tram method using
    `uuencode', which is found on this machine

    Tramp method `sm' confused by extra questions to SSH; try `smx'

    SSH fails; try a telnet method, such as `tm'

        [[Yes, there was a period, after their system administrator
        died unexpectely, when my ISP permit insecure telnet
        connections from anywhere, but prevented secure SSH
        connections -- just the opposite of what had been
        intended...]]



   Maybe the right way to explain it is to start from the user's
   situation.  We should get the user to log in to the remote host from
   the command line.  The command used there will tell us something
   about which methods are applicable.

This is what I have been doing.  That is how I know that one of the
machines I connect to lacks `mimencode'.  This method should be a last
resort, since it is so inefficient.  

It is a way to track down bugs, not a way to get tramp working on 15
different machines, some running non-GNU operating systems, such as
SunOS, as 16 different users.


   Then we could tell the user to issue commands like "type mimencode"
   and "type uuencode", as you have used.  This will tell us more about
   the applicable methods.

This kind of information should be in the `Recovery from Problems'
section.  It is useful for helping people to report bugs. If Tramp is
able to connect, it should be able to run `type mimencode'
automatically if need be.

-- 
    Robert J. Chassell                  bob@rattlesnake.com
    Rattlesnake Enterprises             http://www.rattlesnake.com