Re: BugZilla issue 213, PROPFIND:infinity

Lisa Dusseault wrote:
> I had previously misunderstood what we'd discussed and thought we'd 
> allow PROPFIND depth infinity requests to be rejected IF they posed a 
> performance problem.  Allowing them to be rejected consistently is, IMO, 
> effectively deprecating PROPFIND Depth Infinity, because then clients 
> couldn't rely on it.

Yes, we discussed that very issue. Requiring the server to execute the 
PROPFIND if it happens to be cheap is not only bad spec writing (how 
would you do a compliance test?), it also doesn't make sense (in many 
cases, even counting the descendants of a collection will be almost as 
expensive as returning them -- the recursion is the problem!).

Furthermore, servers today already work this way (they are *configured* 
to reject PROPFIND/infinity in any case), and I'm not aware of any 
client problems because of that.

> That is possibly a fine thing.  If we learned that most clients don't 
> use it anyway, then we could deprecate PROPFIND depth infinity and 
> servers wouldn't even have to worry about implementing it at all.

Best regards, Julian

Received on Saturday, 4 February 2006 09:46:33 UTC