W3C home > Mailing lists > Public > w3c-dist-auth@w3.org > October to December 1997

RE: more v5 draft questions

From: Jim Whitehead <ejw@ics.uci.edu>
Date: Fri, 26 Dec 1997 15:34:35 -0800
Message-ID: <01BD122E.2B80D160.ejw@ics.uci.edu>
To: "'Jim Davis'" <jdavis@parc.xerox.com>, "w3c-dist-auth@w3.org" <w3c-dist-auth@w3.org>
Cc: "Roy Fielding (E-mail)" <fielding@ics.uci.edu>

Thank you for your comments.  Some affirmations and responses below.

On Tuesday, December 16, 1997 5:30 PM, Jim Davis 
[SMTP:jdavis@parc.xerox.com] wrote:
> 8.11.5 has the Depth value "infinity" written as "Infinity".  Should be
> lower case


> Sometimes the opaquelocktoken scheme is written in all lower case 
> and sometimes in mixed case (8.12.5)

This should definitely be converted to all lower case for all instances.

> Also 8.13.10 writes the locktoken "directly", rather than as a Coded-URL,
> as mandated by 9.11

Good catch.

> Perhaps it would be useful to move the definition of Coded-URL from 9.5
> Destroy Header to someplace more generic, since it is used in many places
> (e.g. Enforce-Live-Properties, Lock-Token request and response headers 

Agreed, especially since the Destroy header is being moved to the 
versioning document.

> Why is Coded-URL not used in the "AdditionalLocks" portion of the 
> request header (unless that's because AdditionalLocks is to be dropped, 
> I suspect?)

As you suspected, AdditionalLocks is being worked on.  If it remains, it 
will use the Coded-URL syntax.

> Finally, can someone explain to me why we need "Coded-URL"?  Why can't we
> just write property names as property names?  We don't seem to need to 
> the URI in the Destination header, so why do we need to code it when used
> in e.g. Enforce-Live-Properties, or for a list of locks?

My recollection of the rationale (as originally told by Roy Fielding) is as 
- We needed a list of URLs, which implies a set of URLs with some form of 
- Separator characters are ideally not legal URL characters, since this 
allows URLs to be placed into the protocol stream without an encoding step 
to URL-encode instances of the separator character
- The typical separator character, "," happens to be a legal URL character, 
hence it would be difficult to detect the end of a URL in a comma-separated 
- While space is also not a valid URL character, its use as a separator is 
deprecated (I can't remember why -- perhaps due to gateway manipulation of 
- The characters "<" and ">" are not valid URL characters, and hence, when 
used to wrap a URL, allow the wrapped URL to be included in a 
comma-separated list.
- Since many headers use comma-separated lists, a comma-separated list of 
URLs would be able to use existing parsers in HTTP applications. Use of 
another separator would cause applications to develop a new parser for that 
type of separated field.
- While the quote character could be used as a URL wrapping character, we 
decided to be consistent with RFC 2068 which uses the "<" and ">" 
characters to wrap URLs in the Link header in the appendix.  Granted, this 
is a somewhat weak rationale now that this appendix has been removed.  I do 
recall there was a reason why quote characters aren't the ideal choice, but 
the exact reason escapes me at present (Roy?).

For the Destination header, we follow the conventions of RFC 2068, where 
singleton URLs are not encoded in header fields.  The exception to this is 
the Lock-Token header, where we do use the Coded-URL.  This, at first 
blush, does appear to be arbitrary, and could be changed back to an uncoded 
URL, unless there is a reasonable case that this field could become a list 
of URLs in the future (have to think about this one).

> To take this from a slightly different angle, when a property name is
> written as a coded-URL, it looks just the same as if if were an XML open
> tag.  Would it be better to express them as "empty" tags instead?  The
> difference is that <foo><bar><quux> is not legal XML, but
> <foo/><bar/></quux/> is.  Another advantage, although unlikely to arise 
> practise is that XML allows for tags to have arbitrary attibutes, e.g. I
> could have  <foo type="3"/>.  Here, the "property name" is just foo.  so
> there's a real difference, I think, between expressing this is a 
> of the property name as opposed to as a tag.

Given the context of use of Coded-URLs is limited to headers only, I don't 
see there being a cause for confusion.  Where do you see this confusion 

- Jim
Received on Friday, 26 December 1997 21:47:49 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 20:01:12 UTC