etag or itag?

Whilst reviewing the rfc for delta encoding, I realized that
my understanding of "etag" may be incorrect.

I had assumed that etags were associated with the
"entity contents" of a particular response -- that if any of the bits  of
the body of a response were to change, then so should the etag. 

   Thus, suppose that  foo.txt can be returned as is, or with  gzip    
   content coding: regardless of how it's content encoded, the end 
   result (after ungzipping in the latter case) is identical.  
    Nevertheless, since the actual contents are different (one is
    plain text, the other is a gzip binary), the etags should be 
    different.

I now wonder if this is incorect. 
     Instead, should the etag be associated with the underlying 
    "instance"? 
That is, with the contents of a response BEFORE any content 
codings, OR range selection, have been applied. 

   In terms of the above example, both responses should  use the 
   same etag. 

The latter interpretation (as an "instance tag") makes more
sense to me, even though it complicates server 
implementation. 

-----------------------------------------------------------
Daniel Hellerstein
danielh@crosslink.net
http://www.srehttp.org
-----------------------------------------------------------

Received on Saturday, 18 March 2000 17:32:39 UTC