ange-ftp and sftp?

ange-ftp and sftp?

[Jerry Sievers]

Hello.  This is on Emacs 20.7

Wondering what it would take to have ange-ftp use sftp instead of
regular ftp?

The sftp requires the hostname be given as command line parameter
rather than set using OPEN as per traditional ftp.

Aside from this, not sure what other incompatibilities exist.

Rather learn if this has been addressed before I try wrapping sftp
with something that can emulate what traditional ftp clients expect.

Thanks.

-- 
-------------------------------------------------------------------------------
Jerry Sievers   305 854-3001 (home)     WWW ECommerce Consultant
                305 321-1144 (mobile	http://www.JerrySievers.com/

Re: ange-ftp and sftp?

[Phillip Lord]

>>>>> "Jerry" == Jerry Sievers <jerry@jerrysievers.com> writes:

  Jerry> Hello.  This is on Emacs 20.7

  Jerry> Wondering what it would take to have ange-ftp use sftp
  Jerry> instead of regular ftp?

  Jerry> The sftp requires the hostname be given as command line
  Jerry> parameter rather than set using OPEN as per traditional ftp.

  Jerry> Aside from this, not sure what other incompatibilities exist.

  Jerry> Rather learn if this has been addressed before I try wrapping
  Jerry> sftp with something that can emulate what traditional ftp
  Jerry> clients expect.


It would probably be easier just to use tramp instead. I don't know if
it will do sftp, but I do know that it can tunnel directly through
ssh. I use this all the time. 

I suspect that you have never tried tramp. It's a fine package!

http://savannah.gnu.org/projects/tramp

Phil

Re: ange-ftp and sftp?

[Miles Bader]

Phillip Lord <p.lord@cs.man.ac.uk> writes:
> It would probably be easier just to use tramp instead. I don't know if
> it will do sftp, but I do know that it can tunnel directly through
> ssh. I use this all the time. 
>
> I suspect that you have never tried tramp. It's a fine package!

There was some discussion of this before, and the conclusion reached was
just the opposite.

Tramp works ok in it's domain, and is a valuable tool in situations
where nothing else works, but it's essentially very inefficient (if you
look at what it's doing you'll gag on your wheaties).  Ange-ftp, by
relying on ftp's direct support of various file-system operations, can
be _much_ more efficient (and more reliable) -- and sftp is basically
like ftp with many of the bogosities removed.

Certainly ange-ftp is an old and crufty mode, and probably would need a
bunch of work to eliminate assumptions about ftp, but apparently tramp
would need even more work to handle sftp properly.

-Miles
-- 
I'm beginning to think that life is just one long Yoko Ono album; no rhyme
or reason, just a lot of incoherent shrieks and then it's over.  --Ian Wolff

Re: ange-ftp and sftp?

[Phillip Lord]

>>>>> "Miles" == Miles Bader <miles@gnu.org> writes:

  Miles> Phillip Lord <p.lord@cs.man.ac.uk> writes:
  >> It would probably be easier just to use tramp instead. I don't
  >> know if it will do sftp, but I do know that it can tunnel
  >> directly through ssh. I use this all the time.
  >>
  >> I suspect that you have never tried tramp. It's a fine package!

  Miles> There was some discussion of this before, and the conclusion
  Miles> reached was just the opposite.

  Miles> Tramp works ok in it's domain, and is a valuable tool in
  Miles> situations where nothing else works, but it's essentially
  Miles> very inefficient (if you look at what it's doing you'll gag
  Miles> on your wheaties).  Ange-ftp, by relying on ftp's direct
  Miles> support of various file-system operations, can be _much_ more
  Miles> efficient (and more reliable) -- and sftp is basically like
  Miles> ftp with many of the bogosities removed.

  Miles> Certainly ange-ftp is an old and crufty mode, and probably
  Miles> would need a bunch of work to eliminate assumptions about
  Miles> ftp, but apparently tramp would need even more work to handle
  Miles> sftp properly.


I didn't know this. From my perspective, efficiency is not that much
of an issue though. It would be nice if it ran a little quicker, but
it works well for me. 

I'm guessing that tramp tries to abstract over the different transport
mechanisms, and so can't use their features as much?

Phil

Re: ange-ftp and sftp?

[Miles Bader]

Phillip Lord <p.lord@cs.man.ac.uk> writes:
> I didn't know this. From my perspective, efficiency is not that much
> of an issue though. It would be nice if it ran a little quicker, but
> it works well for me. 

In my experience it can take tramp up to 30-40 seconds to open the first
file on a remote machine, even over a local network!  Ange-ftp isn't
great either, but it's faster than that (especially once the connection
is established).  I expect sftp would be even faster if made to work.

> I'm guessing that tramp tries to abstract over the different transport
> mechanisms, and so can't use their features as much?

Not really -- tramp is written for a very specific method of operation:
it uses a user login session on the remote machine, doing all
communication over the standard input and ouput channel for that login.
I think it's common to have remote login capabilities even when FTP
doesn't work, so tramp might be more widely usable.

Tramp tries to make few assumptions about the availability of various
programs on the remote machine, and doesn't assume a clear
communications channel, so it is forced to use some very clunky and slow
methods (for instance, it actually downloads various scripts to the
remote machine when it first connects, and always encodes files for
transfer using a conservative safe encoding).

-Miles
-- 
"1971 pickup truck; will trade for guns"

Re: ange-ftp and sftp?

[Phillip Lord]

>>>>> "Miles" == Miles Bader <miles@gnu.org> writes:

  Miles> In my experience it can take tramp up to 30-40 seconds to
  Miles> open the first file on a remote machine, even over a local
  Miles> network!  Ange-ftp isn't great either, but it's faster than
  Miles> that (especially once the connection is established).  I
  Miles> expect sftp would be even faster if made to work.

  >> I'm guessing that tramp tries to abstract over the different
  >> transport mechanisms, and so can't use their features as much?

  Miles> Not really -- tramp is written for a very specific method of
  Miles> operation: it uses a user login session on the remote
  Miles> machine, doing all communication over the standard input and
  Miles> ouput channel for that login.  I think it's common to have
  Miles> remote login capabilities even when FTP doesn't work, so
  Miles> tramp might be more widely usable.

Presumably this is only true for inline methods? Things like the scp
method can't assume login capability.


  Miles> Tramp tries to make few assumptions about the availability of
  Miles> various programs on the remote machine, and doesn't assume a
  Miles> clear communications channel, so it is forced to use some
  Miles> very clunky and slow methods (for instance, it actually
  Miles> downloads various scripts to the remote machine when it first
  Miles> connects, and always encodes files for transfer using a
  Miles> conservative safe encoding).


Ah. Yes, you were right in your first post. I really didn't want to
know that as it is, officially, too scary. 

Phil