Do we want to import this change?

[pauamma]

http://community.livejournal.com/changelog/9267963.html

I'm not sure of the implications. Will the extra headers this causes perlbal to add give us better caching, worse, or indifferent?

Comments

[exor674]

As is, nothing -- we'd need a Perlbal (config?) change to spit out the headers too when ?v=1 is present.

[pauamma]

OK. What would be the net result of this combined with appropriate perlbal config changes? (IOW, do we want this or not?)

[exor674]

theoretically, depending on what headers the perlbal changes set, and assuming we rarely change the icons, could reduce load a bunch. Right now we don't serve ANY cache control headers, so telling the browser to keep the static images around could help -- but could also lead to them having old images blah blah -- Does seem like a good idea though.

Might be a better way to do it though, especially since we add a v=xxxxxxxxx to the JS urls, would be to send cache-control headers with everything on s.dreamwidth.org ( and check if there's anything there that we don't want to cache, of course )

[alierak]

Don't browsers keep small static files around by default anyway though? I thought the purpose of the ?v=1 and perlbal hack was so you could force clients to uncache static files more quickly.

[exor674]

Assuming that is right then, JUST adding v=1 to our images will do nothing to improve caching.

[pauamma]

http://bugs.dwscoalition.org/show_bug.cgi?id=3108 (thanks for the explaining)