Re: TRAMP password caching

[Eli Zaretskii]

> From: Michael Albinus <address@hidden>
> Cc: address@hidden,  address@hidden,  address@hidden
> Date: Mon, 04 Dec 2006 21:09:39 +0100
> 
> >> Under w32, it still might be desirable to go back to plink as default
> >> method. Here I would like to get a recommendation from w32 users.
> >
> > What is the issue here, and how does one see it in tramp.el?  (I
> > didn't follow this thread, sorry.)
> 
> Recently, tramp-default-method has been changed to "scp" respective
> "pscp" methods. It has been reported, that people experience a lot of
> password requests from tramp, when they do not run ssh-agent, or when
> password.el is not available.
> 
> For openssh like implementations, I've (tried to) fix(ed) it by adding
> ControlMaster/ControlPath arguments to the "scp" commands. AFAIK,
> these arguments are not available under w32, so it shall be checked
> whether "pscp" is still the preferrable default method.

I checked this now: on Windows, pscp behaves like scp and plink
behaves like ssh on Posix platforms.  In other words, we should indeed
change the default method on w32 to plink, because that is faster and
asks for the passphrase fewer times.

> Of course, one would be on the safe side when PuTTY's pageant is
> running (the ssh-agent pendant). Maybe one could check this during
> initialization, and set tramp-default-method to either "plink" or
> "pscp", depending on the result.

I will try to see if this can be done easily from within a running
Emacs.

Btw, why does pscp method work so hard when pageant is not running?  I
see a lot of activity, including remote shell setup, remote `ls'
command to get a directory, sending Perl scripts to the remote
machine, etc.  Why doesn't it simply invoke pscp to copy the darn
file?  Is it for file-name completion, perhaps? if so, maybe we should
give users a way of disabling completion?