- From: Lisa Dusseault <lisa@xythos.com>
- Date: Sun, 26 Nov 2000 22:48:17 -0800
- To: "Geoffrey M. Clemm" <geoffrey.clemm@rational.com>, <w3c-dist-auth@w3.org>
I agree with everything except... > -----Original Message----- > From: Geoffrey M. Clemm > To: w3c-dist-auth@w3.org > Subject: Depth infinity PROPFIND (was: [RFC2518 Issue] PROPFIND > 'allprop' usage) > > Note: the > proposal is not that you use a 507 to omit some of the properties in a > response element, but rather to omit *all* of the requested properties > in the response element. I disagree about the scope of the proposal, at least the proposal that _I'd_ like to make. I know this isn't exactly the same issue, but servers want to (anthropomorphically-speaking) be able to omit property values which are too expensive to compute when the PROPFIND covers many resources. Thus, with a PROPFIND request covering a number of resources simultaneously, the server could choose to include the 507 response rather than the property value. This is even easier than omitting sub-directories, because 2518 already has examples of returning an error for a specific property (p 25-26 shows two properties omitted due to permissions). It's because these two issues, expensive properties and depth-infinity PROPFIND, are so similar that we've been unable to separate them in mail threads. Luckily I think they can be dealt with in a very similar manner. I'd like it to be explained, in RFC2518 ideally or at least understood through the mailing list, that a client must be prepared to get incomplete PROPFIND results containing 507 codes in status elements to indicate that it was too expensive to calculate certain properties or to recurse into certain trees. With this much information, the client is perfectly able to send more specific requests to fill in the holes, whether they be missing property-values or missing resource-details. This ought not to be too uncomfortable a change for existing clients: since this is legal within 2518, it should not absolutely break them. And, there is nothing forcing servers to return incomplete results, so any server implementor can choose to remain compatible with existing clients relying on depth-infinity PROPFIND returning complete results. Lisa
Received on Monday, 27 November 2000 01:46:06 UTC