W3C home > Mailing lists > Public > w3c-dist-auth@w3.org > April to June 2001

RE: Status code for creating lock-null resource

From: Eric Sedlar <eric.sedlar@oracle.com>
Date: Thu, 28 Jun 2001 11:07:33 -0700
To: "Lisa Dusseault" <lisa@xythos.com>, "Alan Kent" <ajk@mds.rmit.edu.au>, <w3c-dist-auth@w3.org>
Message-ID: <NDBBKNOGFKEBJOOOIOOLIEMACBAA.eric.sedlar@oracle.com>
I think the historical view that resulted in LNR's is as follows:

* There are valid scenarios where clients want to create a resource and
	lock atomically, or create a collection and lock atomically
* We can't handle this for regular resources since we can't change PUT,
	although we could solve this for MKCOL

My suggestion is built on Lisa's suggestion:

* Make LOCK create empty resources (with length 0) that are real resources
	if the URL is not mapped and a special header is specified
	(say Create: true, kind of like the O_CREAT flag on open() on Unix)

* Make MKCOL take a header specifying that a lock should be granted as
	well, like Lock: true


I've always had a viceral negative reaction to creating gratuitous empty
files, but on retrospect I think this comes from working on versioning
systems where this would create a useless version that the customer doesn't
want.  In this case, since a deltaV content creation client would probably
do:

* Open -->   LOCK / Create: true
* Save -->   PUT followed by VERSION-CONTROL followed by UNLOCK

so there is no problem of an extraneous version.

--Eric
Received on Thursday, 28 June 2001 14:03:46 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 2 June 2009 18:43:56 GMT