RE: ACTION-550: Draft some initial material for Section 2.3 of the Guidelines

I appreciate the work done by Sean Patterson. However I think a lot of 
the assumptions made in there is questionable at best. I'll write a 
comment on that later. But first, I have something to say about the 
technical suggestions by the Sean and Jo.

Debate of whether the transcoding proxy can alter the User-Agent header 
aside, I think the suggestion by Jo is problematic. What I mean is that 
in order to avoid introducing a new HTTP header, you are suggesting a 
mechanism where the transcoding proxy makes a second request with the 
original User-Agent. While HTTP itself is a stateless protocol, in 
practice, however, making the same request twice will be problematic in 
many cases in practice. What if the request is a forum post? What if the 
request is an image request for an ad, with which each request is 
counted as an impression and therefore, money? Not to mention that it 
introduces needless overhead to the whole request-response chain from 
the browser to the origin server. What if you have multiple transcoding 
servers in the delivery chain? I'd rather that a new HTTP header be 
introduced, and that there be no second requests on behalf of the 
browser in this case.

I think it is a good idea for servers to always return a Vary: 
User-Agent or Vary: * . Shouldn't this be in the Mobile Web Best 
Practices? I can see in 5.4.15 that providing caching information is 
recommended. Perhaps it needs to be spelled out that the server needs to 
return the Vary: header as well, if the server does use the User-Agent 
for content negotiation. Well a tad off-topic...

Regards,
Nigel.

Received on Friday, 28 September 2007 00:54:07 UTC