Re: PROPOSAL: Weak Validator definition [i101]

On Sat, 2008-03-15 at 22:40 +0100, Werner Baumann wrote:
> Henrik Nordstrom wrote:
> > Quite likely there is many examples out there in the wild along the
> > lines of the pageviews counter example found in the RFC, but semantic
> > equivalence fits within "good enough from the servers point of view".
> 
> I am not sure whether this is meant by you the way I understand it: Do 
> you mean "good enough from the servers point of view" is o.k., because 
> it includes your understanding of semantic equivalence? Or is it meant 
> the other way round?

Basically that I am ok with the change.

> To be pedantic about pageview counters (what I ain't): a change in the 
> pageview count is a semantic change. But I agree, it would be a good 
> example for weak etags, because it is not important. I, personally, 
> would accept strong etags in this case (although this violates the spec).

strong ETags MUST NOT be used in that case as the object isn't binary
equivalent. A property of strong ETags is that it's possible to merge
ranges of the object received in different responses, and this would
result in garbage if the same ETag is shared among two slightly
different objects (except in very rare cases).

> To be honest: I would not know, how to get this working with Apache. 
> Apache's default etag-handling can not deal with this kind of weak 
> etags.

Apaches default etag handling can note deal with weak etags. But any
CGI/PHP script or similar you use for generating content can do so just
fine.

> You would have to override apache's etag-handling and do it all 
> by your own. Quite a lot of effort for a toy like pageview counters.

Change the example to "visitor counter" and it may make more sense.

Regards
Henrik

Received on Tuesday, 18 March 2008 13:29:27 UTC