Re: 14.36

Re objections raised in:

> http://www-uk.hpl.hp.com/people/ange/archives/http-wg-archive/1012.html
and
> http://www-uk.hpl.hp.com/people/ange/archives/http-wg-archive/11196.html

Ted Hardie added:

> I think the current design risks semantic transparency in a
> small number of cases to reduce round trips in an equally small number
> of cases.

Perhaps you meant "semantic consistency", since "transparency" applies
only to things that you (look) through, such as a proxy cache.

However, in none of the designs considered lack of semantic
consistency: whatever the server sends, it's labelled as to what it
is, and, after all, this is not an issue of cache transparency.

The only question is how a server should respond to a request
indicates a different indication of server length from client's
expected length.  Any response at all that is self-consistent actual
length of the object would do. However, we remembered discussion of a
number of situations where it was *most* useful for the server to
return the maximum amount of information consistent with the client's
request.

I don't know how the wording could be changed to make this clearer,
though, or if it needs to.

- Larry