- From: Larry Masinter <masinter@parc.xerox.com>
- Date: Tue, 12 Nov 1996 23:55:11 PST
- To: ben@algroup.co.uk
- CC: kt@nttlabs.com, w3c-dist-auth@w3.org
# The possibility that the retrieval method might change is worrying, though. Should the Link attribute probably should expire at the same time that the data expires? One of the differences between the web and traditional file systems is that the web is willing to deliver stale data, and clients are expected to somehow cope. This is an important feature of the web, and I don't think we can presume that it won't be there while we're doing DAV. The general principal is that GETs are optimistic, and PUTs and POSTs are done with validators. That is, you go ahead and GET and read whatever you want, but when you go to PUT or POST, you do it with a guard against the entity tag. There are no transactions, so all atomic operations must be accomplished with a single POST. This is pretty reliable, actually, and doesn't require us to muck up web transaction semantics just to add transactions. Larry
Received on Wednesday, 13 November 1996 03:02:02 UTC