sftp url rfc draft

[Salowey, Joe]

Hi Robey,

I see the confusion and I share the confusion.   I looks the URI RFC
(3586) allows this in the text portion, but the ABNF looks like it may
disallow a double slash.  In general the two URLs have different meaning
since the use of the "/" indicates it is part of the URL hierarchical
structure and the use of the "%2F" indicates it is part of the path
segment. It seems that the "/" representing the root of the file system
is hierarchical in nature and would not be escaped (the current draft
would be wrong then).  I will have to check with a URL expert on this.

What packages implement the de-facto standard of absolute paths only?
Is this for SFTP protocol?  If this is the predominant way that the URI
is in use then it might be a good idea to have the document reflect this
convention.  

Cheers,

Joe

> -----Original Message-----
> From: Robey Pointer [mailto:robey at lag.net] 
> Sent: Sunday, December 11, 2005 1:45 PM
> To: Salowey, Joe; suehring at braingia.com
> Cc: Bazaar-NG BZR
> Subject: sftp url rfc draft
> 
> Hi there!
> 
> At the bzr project (www.bazaar-ng.org) we plan to use sftp 
> urls to access remote source-code repositories, and there's 
> been some discussion about whether to use the current 
> de-facto standard (absolute paths only) or the new RFC draft 
> that you two have posted
> here:
> 	
> <http://www.ietf.org/internet-drafts/draft-ietf-secsh-scp-sftp-ssh-
> uri-03.txt>
> 
> which supports both absolute and relative paths.
> 
> One point that's come up is that nobody wants to type "%2F" 
> in an url to get the equivalent of the old url behavior.  The 
> RFC seems to sometimes imply that using "//" would be okay 
> (since it would be otherwise meaningless), but sometimes 
> implies that this would be unacceptable.
> 
> Because this point could affect our decision of which format 
> to use, could you clarify if a path like
> 
> 	sftp://user@host//absolute/path
> 
> would be an acceptable equivalent to
> 
> 	sftp://user@host/%2Fabsolute/path
> 
> ?
> 
> Thanks!
> 
> robey
>