On Sunday 25 August 2002 14.46, Robert Collins wrote:
> I propose that we make 2.6 a very small change over 2.5, given that
> it already has the reworked comms code from Adrian.
Agreed.
> I'd like to see the following in 2.6:
> * Some authentication tidyups I intend to do Real Soon Now. (to
> support delay pools w/ all auth types).
Don't forget all the other access lists.. I think the auth identity 
should be cached in the request where appropriate.. I have no problem 
with proxy_auth being enforced in http_access if to be used in other 
access lists, but I have a problem with proxy_auth not working on 
other access lists even if already successfully processed in 
http_access for the same request.
> * rproxy (if appropriate - Henrik?)
I'll try to get parts of it over at least. Cannot promise all of it 
gets there. There is some important plumbing and fixes in the rproxy 
branch that really should get into HEAD/2.6 even if rproxy itself 
does not..
> * the client side streams I've done for ESI.
OK.
> * Optionally, ESI itself.
OK.
> * Optionally, any other sponsored work that needs to get into
> mainline sooner rather than later.
I have ETag caching.. think we need a little discussion on this topic. 
The implementation in the etag branch (etag-2_5.. the head etag 
branch is not yet 100% updated with the store client changes) is 
working quite reasonable, but it might need a bit of work to apply to 
your ESI client side reorganisation.
Some discussion on the current etag branch implementation can be found 
at http://devel.squid-cache.org/etag/. Basically it operates by 
saving a secondary index of "variants" on the StoreObject of the URL 
when detecting Vary headers, and storing the actual objects under 
separate keys (ETag driven, or Vary header values if no ETag is 
present). Used successfully in production in an environment with lots 
of document variants by the project sponsor.
> * If range caching is a showstopper, reinstated range caching.
Some kind of Range processing should be there in a STABLE release. At 
least Range processing of cache hits.
> Are there any other near-to-completion projects on
> devel.squid-cache.org that folk would like to see in 2.6?
I have one not yet published (awaiting sponsor).
Regards
Henrik
Received on Sun Aug 25 2002 - 15:21:27 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:16:08 MST