Re: Unify Ange-FTP and Tramp filename formats?

[Steve Youngs]

|--==> "KG" == Kai Gro <address@hidden> writes:

  KG> (1) Turn off EFS completely, so that it is never invoked.  This means
  KG>     there will be no conflicts between Tramp and EFS filenames, as
  KG>     there are no EFS filenames.

This assumes that the user has the Tramp package installed.  Because,
unlike GNU/Emacs, our packages are optional.  So turning off EFS is a
bad thing.  Unless of course you do the "switching off" from within
Tramp. 

  KG>     The general Tramp syntax will be /address@hidden:file.

  KG> (2) Tell Tramp to invoke EFS for certain filenames.  For instance, if
  KG>     method=ftp, EFS is always invoked.

  KG> (3) Allow user to specify default method based on user/host
  KG>     combination.  For instance, if user=anonymous, EFS could be
  KG>     invoked.

So you're going to make the user add "[ftp]" or "anonymous@" to their
filenames, only to strip it off to pass it on to EFS.  Kai, that
simply doesn't make any sense.

  KG> One way to make sure the XEmacs Package tools work is to switch them
  KG> from the /address@hidden:file format to /address@hidden:file (literal
  KG> square brackets).

No that won't fly.  That makes Tramp a requirement for using the
Package tools.  

I would veto any proposal to modify the Package tools in such a way
that means another package must be pre-installed before they can be
used.  We already require 2 packages (3 if you're running Mule) be
pre-installed.  That's a big failing of the Package system and we want
to reduce that requirement to nil.

  KG> Another way to make sure the tools work is to add an entry to
  KG> tramp-default-method-alist that user=anonymous implies ftp.  Then
  KG> filenames of the form /address@hidden:file will continue to work with
  KG> EFS (I presume that the filenames used by EFS are of this form).

Yes, but '/host:file' implies '/address@hidden:file' and I think that
most people would do their anonymous EFS'ing with '/host:file'.  And
to counter my own argument, you could easily overcome this with a
simple '(or ...)' or a '(cond ...)'.

  KG> So, I think it's clear now that "2 of those will break EFS" does not
  KG> apply.

Perhaps.  But it could potentially confuse users for no good reason.
AFAICT (I followed the discussion on emacs-devel) there isn't any
point to this.  There just doesn't seem to be any pluses that make it
worth the effort to implement.

  KG> When I first read Richard's suggestion, I thought it was a really
  KG> horrible idea.  But the more I think about it the better I like it.

For me, the more I think about it, the more horrible it gets.  But
Tramp isn't my package so I'll shutup now. :-)

-- 
|---<Steve Youngs>---------------<GnuPG KeyID: 10D5C9C5>---|
|            XEmacs - It's not just an editor.             |
|                    It's a way of life.                   |
|------------------------------------<address@hidden>---|