|
Hi James,
On 06/09/2015 03:59 PM, James Smith wrote:
In many cases it will only be a few packets anyway so won't
actually make that much difference!
The point is that it is better to stop the request in the first
place by setting the appropriate expires/cache control header...
than use the etag mechanism...
In case it wasn't clear, we very much agree on this one, my question
was centered on the ETag in particular.
Thanks for your reply!
James
On 09/06/2015 14:56, Frederik Nosi
wrote:
Hi James,
On 06/09/2015 02:36 PM, James Smith wrote:
Yes - it is the request over head - the client will still make
the request at which point the server has got to decide has it
changed before even - which for most static requests is the
heaviest (slowest) part before returning the not-changed
response - and then serving the content!
But at this point the server in case of a positive match will
send just a 304 reply with no content, thus saving bandwith and
time (due to eventual roundtrips) no?
You are better to:
(a) set near future or mid future headers [ expires in a month
or in a year]
Sure, the best request is the one that does not even come :-)
(b) alter filenames if you significantly change the file
contents [ we use MD5 of content for js/css ]
This only if you're in the posisition to decide the site layout
though.
Note this is "hyper-tuning" of Apache... some people may want
to enable it - it was originally set up when most users were
on 28K/33.6K modems (or slower) and the transfer of data was
the slow part of the equation!
James
[...]
Thanks,
Frederik
-- The Wellcome Trust Sanger Institute is operated by Genome
Research Limited, a charity registered in England with number
1021457 and a company registered in England with number 2742969,
whose registered office is 215 Euston Road, London, NW1 2BE.
|
