RE: I-D ACTION:draft-ietf-secsh-scp-sftp-ssh-uri-00.txt

["Joseph Salowey" <jsalowey%cisco.com@localhost>]

URI's have a pretty specific format detailed in RFC 2396.  Here is my
interpretation

For SCP/SFTP we are defining Abolute URI

AbsoluteURI = scheme ":" (hier_part | opaque_part)

If we deicide that SCP URIs are hierarchical in nature and will be
parsed by a URI parser then we use a hier_part.

Hier-Part = "//authority [abs_path]
Abs_path = "/" path_segments

In this case we use the standard URI definitions to define the authority
(User@host:port).  Along with this comes the leading '/' and the use of
'/' as a path separator.  This means that SCP will have to have a
minimal URI paser.  This means the C:\mywindows\files may be represented
as

Scp://host/c:/mywindows/files

If we choose to make them opaque then what comes after scp: is defined
and parsed by SCP then the URI for the above file might look like (other
parsing schemes are possible, the point is they do not have to adhere to
URI parsing).

Scp:user:@host::c:\mywindows\files

In general since filesystems have a hierachical structure it is probably
better to go with hierarchical naming.  This makes it possible to use
relative URLs and reference a resource in different ways.

Thanks,

Joe

> -----Original Message-----
> From: ietf-ssh-owner%NetBSD.org@localhost 
> [mailto:ietf-ssh-owner%NetBSD.org@localhost] On Behalf Of Simon Tatham
> Sent: Wednesday, August 13, 2003 4:26 PM
> To: ietf-ssh%NetBSD.org@localhost
> Subject: Re: I-D ACTION:draft-ietf-secsh-scp-sftp-ssh-uri-00.txt
> 
> 
> Joseph Salowey <jsalowey%cisco.com@localhost> wrote:
> > So how about
> > scp://hostname/c:/bin/thingy
> 
> The difficulty with that is that you really want the SCP 
> client to send the pathname 'c:/bin/thingy', without a 
> leading slash. But when you say `scp://hostname/usr/bin/foo', 
> you want the SCP client to send `/usr/bin/foo', _with_ the 
> leading slash. So this design requires the _client_ to do 
> something clever about conditionally stripping the leading 
> slash. Surely?
> 
> (Unless you're mandating that the _server_ in the 'c:/' case 
> must be capable of throwing away a leading slash? But that 
> seems like a requirement on server behaviour as well as URI 
> format, which is fairly serious remit creep...)
> 
> -- 
> Simon Tatham         "I'm going to pull his head off. Ear by ear."
> <anakin%pobox.com@localhost>                          - a games teacher
>