Re: ETags?

Julian Reschke wrote:

> Elias Sinderson wrote:
>> So, in effect, the spec will state that dead properties MUST be the 
>> same across all bindings to a given resource, and that (by remaining 
>> silent on the issue) live properties MAY vary depending on server 
>> implementation. [...]
> I'm not happy with this text, as it sort-of *encourages* people to 
> define live properties that way.

Just to be clear, I wasn't proposing any specific text, but merely 
observing that a note to client implementors would probably be useful. I 
agree that the use of MAY as above would likely encourage the definition 
of live properties that vary across bindings.

> So *if* the spec makes a statement about dead properties, but 
> *doesn't* make statement about live properties, that should be clear 
> enough ("if the spec author's would have wanted to make a statement 
> about live properties, they would have done it here").

Hmm, I can imagine client implementors having discussions about this. 
Including a single sentence which states that clients can't necessarily 
depend on live properties being the same on different bindings to a 
given resource.

> Having the choice between two text versions that technically say the 
> same, I'm voting for the shorter and simpler one.

I generally agree, however this seems to be a case in which a little 
guidance could prevent a lot of headaches since the question will 
undoubtedly arise. OTOH, it does run dangerously close to Geoffs' 
statement regarding such guidances... I certainly don't feel strongly 
enough about the issue to hold up the BIND spec.


Received on Friday, 21 January 2005 22:36:42 UTC