[squid-dev] Cache-Groups RFC draft

[Alex Rousskov]

On 2024-10-30 08:38, Amos Jeffries wrote:
> On 31/10/24 00:09, Francesco Chemolli wrote:
>> Hi all,
>>    do we have any opinion on
>> https://www.ietf.org/archive/id/draft-ietf-httpbis-cache-groups-02.html
>> ?
> 
> This is a further extension of Mark's variant Key header extensions to 
> HTTP.
> 
> When I proposed adding support for the first steps of these extensions 
> Alex complained that Squid's storage needed major overhaul before it was 
> worth doing.

I hope my assessment (referred to as "complaint" in the above paragraph) 
was correct.


> FYI; The "group" identifier in this HTTPbis Draft is akin to StoreID for 
> a "Vary marker" store object.
> 
> To support this mechanism cleanly Squid should first refactor "Vary 
> marker" to to have multiple headers determining further lookups objects, 
> not just the one Vary pattern.

FWIW, I doubt that group invalidation should be implemented using 
something similar to Vary marker objects that Squid uses today. I 
suspect there are better ways to instantly invalidate large sets of 
cached entries. I assume that we do not need to debate the design (and 
its prerequisites) now, but if somebody decides to actually proceed with 
this project, they should keep this squid-dev thread in mind.


> That is similar matching logic needed for the prior Key (also Vary-Key) 
> header mechanism which Alex has already vetoed as "too complex" to 
> implement.

Implementation complexity is an important factor, of course, but I doubt 
that I have vetoed any mechanisms just because they are "too complex to 
implement".

Alex.