hit metering as 'Proposed Standard' (revised)

Sigh, I think I've run out of time to use 'jet lag' as my excuse, but
being off net for a week really did leave me with too much mail; I
thought I'd reviewed all the messages so far on hit metering, but
clearly missed most of them when I sent my 'summary'.

So, here's a revised summary of the issue:

I think we're having a good discussion about the proposal. I think
there have been several comments and we should expect to see a revised
draft with clarifications and changes that remove some of the

My current reading of the situation:

There is significant interest in forwarding this draft along standards
track. There are serious concerns about its applicability ("does it
give providers enough data") and necessity ("could providers do the
same thing by tweaking max-age") and alternatives ("could we add
headers for asking for statistical sampling rather than hit

Even after revising the draft to meet the other objections, I think we
still have choices:

A: release the specification as 'Experimental' as a way of
   encouraging people in the community to gather data 
B: release the specification as 'Proposed Standard' but
   with an applicability statement that does not encourage
   its general deployment without additional data and investigation
   of alternatives or
C: Wait until those vendors whose customers are eager to deploy
   this solution to supply some data that will convince
   us that the proposal is (a) useful to enough content providers
   (b) cannot be accomplished as easily using current HTTP
   and (c) better than the alternatives.

I don't think we need any more testimonials about who does or doesn't
believe this proposal or want it or think it's important.  We need
facts with data based on actual surveys and measurements.



Received on Tuesday, 26 November 1996 06:21:43 UTC