Re: [Monotone-devel] Re: MTN-over-SSH's syntax

[Daniel Carosone]

On Wed, Aug 02, 2006 at 02:50:26PM -0700, Graydon Hoare wrote:
> Richard Levitte - VMS Whacker wrote:
> 
> >Uhmm, no, scp doesn't use a broken URI syntax, it has its own, which
> >doesn't start with ssh://.  I don't think it's wise to confuse people
> >with something that looks like a URI but isn't in such a subtle way.
> 
> Oops, yes, of course.

Thing is, if I understand the URI format correctly, absolute ssh paths
should actually look like ssh://foo.bar.net//path/to/foo.mtn

After the hostname, the first slash is part of the URI format, and the
second is part of the path locator that makes it absolute.  Locators
for ssh:// URI's (or, at least elsewhere, scp:// URI's) are relative
to the user home dir by default because that's how ssh (or, at least,
scp) behaves.

So if I've understood correctly, our current form,
ssh://foo.bar.net/dir/foo.mtn should instead be relative. The
following forms should also be expected to work:

  ssh://host/~/dir/foo.mtn form, as njs proposed, and
  ssh://address@hidden/~user2/dir/foo.mtn 

But I don't think this needs to be something that we treat as
particularly special or awkward. Looking at the first / as terminating
the user:address@hidden:port part, and the next character as the start
of a local path, this can be seen clearly as just ~ expansion in a
local path the same way a leading / makes the path absolute.  Using a
plain ~ without a username is redundant but harmless (as well as being
consistent with Hg or whichever other system njs mentioned that had
picked this form, sorry I forget which).

Now, that's what I gather from discussions and some usage elsewhere.
I readily admit it's no more or less confusing than some of the
alternatives, until/unless explained carefully.    

--
Dan.

pgpBAwhCh3iQ2.pgp
Description: PGP signature