multi-hop vs proxy

[Michael Albinus]

Hello,

Tramp uses today a syntax describing the complete route to hosts which
could not be accessed directly. I believe this is not very convenient,
because one has always to type the whole route, something like this:
"/multi:rsh:address@hidden:telnet:address@hidden:/path/to.file".

I propose to use a notation of proxies instead. Proxies describe, in
Tramp syntax, the route one has to follow for a given target.

Let's say you declare the following:

(add-to-list 'tramp-default-proxies-alist
             (list "\\`real\\.host\\'" "kai" "rsh:address@hidden"))

Then "/telnet:address@hidden:/path/to.file" would always go via
"/rsh:address@hidden".

Even more than one hop could be defined this way:

(add-to-list 'tramp-default-proxies-alist
             (list "\\`second\\.hop\\'" "" "ssh:first.hop"))
(add-to-list 'tramp-default-proxies-alist
             (list "\\`real\\.host\\'" "" "ssh:second.hop"))

"/telnet:address@hidden:/path/to.file" would be equivalent to (todays
syntax)
"/multi:ssh:first.hop:ssh:second.hop:telnet:address@hidden:/path/to.file".

Advantages are user convenience, code cleanup (multi-hop is a
parameter for most of the Tramp functions, although almost never
needed), and preparation for other Tramp filename syntax like URLs.

Additionally, one could distinguish in "transit" methods and
"endpoint" methods. All methods used today for single-hop are
"endpoint" methods, and all methods defined today for multi-hops are
"transit methods". This would enlarge possibilities for multi-hops,
because out-of-band and foreign methods are currently not allowed for
multi-hop.

What do people think about it?

Best regards, Michael.